Mobile Visualisation Techniques for Large Datasets · 2014-10-31 · techniques for large geographic and categorical datasets. The goals of this research are threefold, these are:

Mobile Visualisation Techniques forLarge Datasets

Submitted in partial fulfilment

of the requirements of the degree of

Bachelor of Science (Honours)

of Rhodes University

Motebang Lebusa

Grahamstown, South Africa

October 2014

Abstract

Visualisations are a quick and powerful tool for data analysis and information acquisition.

The main aim of this thesis was to investigate visualisation and interaction techniques

appropriate for the mobile platform. The mobile platform is investigated, focusing on var-

ious factors that can affect visualisations on mobile devices. The purpose of this research

is to determine if mobile devices be used for visualising geographical and categorical data

for large datasets. To achieve this purpose, three goals were developed, these are i) to

review mobile visualisation and interaction techniques for large datasets, ii) to develop vi-

sualisations based on the findings from related work, and iii) to evaluate the visualisations

for effectiveness and usefulness amongst other factors.

Limited resources were found to be a major hindrance to developing effective visualisations

on the mobile platform. However, advances in the computing field and in the mobile

platform provide an opportunity to develop effective visualisations for the mobile platform.

Results from the literature were used to influence the design and development of the

mobile visualisation application. The application was used to visualise geographical and

categorical data as set out in the research question. An evaluation of the application was

conducted with the purpose of establishing the usability of the visualisation and what

users experienced while interacting with the visualisation. The results demonstrated that

it is possible to develop visualisations for the mobile platform for large datasets. It was

also demonstrated that the visualisations can be intuitive and simple for users to use

without a myriad of complex layouts, fonts and graphics. Also, users reported a level of

satisfaction in using the visualisations and indicated that it was informative to use and

had a simple and easy to use layout.

ACM Computing Classification System Classification

Thesis classification under the ACM Computing Classification System (2012 version, valid

through 2014) (ACM, 2014):

D.2.2 [Design Tools and Techniques]: User Interfaces

H.5.2 [Information Interfaces and Presentation]: Interaction styles

General-Terms: Mobile Visualisations, User interfaces, Interaction styles, Design

Acknowledgements

I dedicate this thesis to my family and friends who have been with me throughout the

difficult year. It would have been a lonely journey to get to this point without the

emotional support and strength they consistently provided.

To my supervisors, Prof. Hannah Thinyane and Mrs. Ingrid Sieborger, I would like to

extend my heartfelt gratitude for the invaluable guidance and direction they relentlessly

gave throughout the course of the research. They helped me pull through the challenging

times. They put up with me when I could not even put up with myself and encouraged

me to rise above my mediocre self - for that, I am grateful.

Without the ingenious input of Mathe Maema my research journey would have been very

murky. She always managed to shed her words of wisdom which brought me back on

course every time (which was literally all the time) I derailed.

I would also like to thank all students who were involved in the user evaluation in one

way or another - their time and input were invaluable. It would not have been possible

to complete my thesis without their feedback. To all my classmates, thank you for the

great year - you have all been great.

This work was undertaken in the Distributed Multimedia CoE at Rhodes University, with

financial support from Telkom SA, Tellabs, Genband, Easttel, Bright Ideas 39, THRIP

and NRF SA (TP13070820716). The authors acknowledge that opinions, findings and

conclusions or recommendations expressed here are those of the author(s) and that none

of the above mentioned sponsors accept liability whatsoever in this regard.

Contents

1 Introduction 9

1.1 Background and Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.3 Research Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.4 Delimitation of Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.5 Structure of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Related Work 12

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.2 Computer visualisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.2.1 Hardware and software specifications . . . . . . . . . . . . . . . . . 14

2.2.2 Design considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.2.3 Categories of visualisations . . . . . . . . . . . . . . . . . . . . . . . 18

2.3 Visualisations in the mobile environment . . . . . . . . . . . . . . . . . . . 18

2.3.1 Hardware and software specifications . . . . . . . . . . . . . . . . . 19

2.3.2 Design considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.3.3 Categories of visualisation . . . . . . . . . . . . . . . . . . . . . . . 24

2

CONTENTS 3

2.4 Visualisation data types . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.5 Interaction techniques for mobile devices . . . . . . . . . . . . . . . . . . . 27

2.6 User tasks in visualisation . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3 Methodology, Design and Implementation 32

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.2 Spiral Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.3 Iteration 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.3.1 Identify . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.3.2 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.3.3 Develop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

3.3.4 Analyse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.4 Iteration 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.4.1 Identify . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.4.2 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.4.3 Develop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.4.4 Analyse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4 User Evaluation 50

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.2.1 Usability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

CONTENTS 4

4.2.2 Usability Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.2.3 User Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4.3.1 Usability Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4.3.2 User Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5 Conclusion 65

5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

5.2 Research Goals Revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

5.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

References 66

A User Evaluation 71

A.1 Information Sheet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

A.2 Consent Form signed by Participants . . . . . . . . . . . . . . . . . . . . . 72

A.3 Evaluation Questionnaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

B Useful Information 76

B.1 Code Listing for Point in Polygon method . . . . . . . . . . . . . . . . . . 76

B.2 Listing for all suburbs used in the visualisation . . . . . . . . . . . . . . . . 76

B.3 Listing for categories used in the visualisation . . . . . . . . . . . . . . . . 77

C Accompanying CD-ROM 78

List of Figures

2.1 (a) The zoomed part gives a detailed information (Baudisch, Good, &

Stewart, 2001) (b) Fisheye view shows more detail at the centre - the focus

(Ross, 2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.2 (a) Halo with arcs referencing off-screen places (Baudisch & Rosenholtz,

2003). (b) ZoneZoom segments a screen into nine views (Robbins, Cutrell,

Sarin, & Horvitz, 2004). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.3 (a) News24 elections app displaying 2014 results by provinces (map-based)

and total number of votes (integer data) political parties got (categorical

data) (News24, 2014). (b) 3D land surface visualisation of wind vector

profiles (Wang, Huynh, & Williamson, 2013) . . . . . . . . . . . . . . . . . 26

2.4 Parallel coordinates visualisation for automobile data (Heer, Bostock, &

Ogievetsky, 2010). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.5 Visualisation of technology stocks representing temporal data (Heer et al.,

2010). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.6 Tree visualisation (Heer et al., 2010). . . . . . . . . . . . . . . . . . . . . . 29

2.7 Network visualisation (Heer et al., 2010). . . . . . . . . . . . . . . . . . . . 29

2.8 (a) User tilts a phone to select letters as he types text (Partridge, Chat-

terjee, Sazawal, Borriello, & Want, 2002). (b) Augmented visualisation

(Soros, Seichter, Rautek, & Groller, 2011) . . . . . . . . . . . . . . . . . . 30

3.1 The spiral model (Sayyed, 2012) . . . . . . . . . . . . . . . . . . . . . . . . 33

5

LIST OF FIGURES 6

3.2 (a) Screen for displaying results for all suburbs b) screen displaying results

for one suburb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

3.3 (a) Relationships in the visualisation application tables . . . . . . . . . . . 40

3.4 System architecture for the entire visualisation application . . . . . . . . . 42

3.5 (a) Screen displays features of the visualisation (b) Screen displaying overall

results for the past six days (c) Screen displaying results viewed by categories 43

3.6 (a) Screen for displaying results for a suburb (b) Displaying results for the

past two days (c) Displaying results for the past six days . . . . . . . . . . 44

3.7 (a) Screen displaying a default suburb (b) Screen displaying a zoom view

of the suburb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

List of Tables

3.1 Sample responses from Vukani suburb . . . . . . . . . . . . . . . . . . . . . 35

4.1 Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.2 Scores for Results of the Tasks . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.3 Question 2 - Usability Feedback . . . . . . . . . . . . . . . . . . . . . . . . 54

4.4 Responses for Question 2 (i) - Usability Feedback . . . . . . . . . . . . . . 55

4.5 Responses for Question 2 (ii) - Usability Feedback . . . . . . . . . . . . . . 56

4.6 Responses for Question 2 (iii) - Usability Feedback . . . . . . . . . . . . . 57

4.7 Responses for Question 2 (iv) - Usability Feedback . . . . . . . . . . . . . 57

4.8 Responses for Question 2 (v) - Usability Feedback . . . . . . . . . . . . . . 58

4.9 Responses for Question 2 (vi) - Usability Feedback . . . . . . . . . . . . . 58

4.10 Responses for Question 2 (vii) - Usability Feedback . . . . . . . . . . . . . 59

4.11 Question 3 - User Experience . . . . . . . . . . . . . . . . . . . . . . . . . 60

4.12 Results from Question 3 - User Experience . . . . . . . . . . . . . . . . . . 60

7

Glossary of Terms

API Application Programming Interface

CPU Central Processing Unit

CRUD Create, Read, Update and Delete

GPU Graphics Processing Unit

HTTP Hypertext Transfer Protocol

ICT Information and Communication Technology

IDE Integrated Development Environment

IO Input/output

IS Information System

JSON JavaScript Object Notation

RAM Random Access Memory

SDK Software Development Kit

SQL Structure Query Language

UX User Experience

WYSIWYG What You See Is What You Get

XML Extensible Markup Language

8

Chapter 1

Introduction

1.1 Background and Context

Visualisations are useful and powerful aids to data analysis and information acquisition

(Ware, 2013). In recent years, the global and regional penetration of mobile devices has

been increasing gradually (GSMA, 2013; ITU, 2014). Advances in the computing field and

mobile platform, combined with the power of visualisations and interaction techniques for

mobile phones, provide mobile users with an opportunity to access and acquire information

and manipulate data at a glance. Providing visualisations on the mobile platform could

prove to be useful and successful in reaching many users due to the increasing penetration

of mobile devices.

The visualisations for this research form part of a larger project called MobiSAM. Mo-

biSAM is a pilot project implemented in the Makana Municipality to support the local

community to participate in local government using mobile phones. Through MobiSAM’s

website1, registered users report problems related to service delivery by taking polls avail-

able on the project website (Thinyane & Coulson, 2012). These polls generate data for the

MobiSAM project. This reporting forms part of MobiSAM’s overarching goal of facilitat-

ing local communities to engage in local government in a meaningful and evidence-based

manner (Thinyane & Coulson, 2012). The current visualisations for the pilot project are

provided through a web interface. Mobile platforms were also targeted for extending the

MobiSAM functionality to the communities (Thinyane & Coulson, 2012). As such, this

research project extends work done on MobiSAM. For this research project, data used

1Available at http://www.mobisam.net

9

1.2. PROBLEM STATEMENT 10

will be sample data for monitoring water quality, from which responses describe water

problems with respect to colour, taste, pressure or smell.

1.2 Problem Statement

Mobile applications need to be developed in order to provide information to mobile device

users. Visualisations have been used for a long time to enhance the amount of information

and insight that can be gained from data (Dix, Finley, Abowd, & Beale, 2004). Given that

the computing resources in mobile devices are limited, developing applications, including

visualisations, for the mobile platform poses a challenge to developers (Van Tonder &

Wesson, 2008). The main statement of purpose for this research is, ‘Can mobile devices

be used for visualising geographical and categorical data for large datasets?’ The large

datasets are envisaged to be generated over years as MobiSAM gains popularity as the

project is in its pilot phase at the time of writing this thesis. Data used in this research

constitutes geographical, categorical and temporal data. The former two data types will

be the main focus for this research work. As an attempt to answer the research question,

three goals have been developed. The next section introduces the three research goals for

this project.

1.3 Research Goals

The visualisation application in this paper attempts to show appropriate visualisation

techniques for large geographic and categorical datasets. The goals of this research are

threefold, these are:

1. review mobile visualisation and interactions techniques for large datasets by finding

state-of-the-art mobile visualisations in this field,

2. develop visualisations based on the findings from related work, and

3. evaluate the visualisations for usability and user experience.

1.4. DELIMITATION OF SCOPE 11

1.4 Delimitation of Scope

The application will be developed for Android, specifically version 4.0 and higher. De-

velopment tools to support this include the Eclipse Android Developer Tool (ADT) used

primarily for development on the Android platform. The Google Maps API will be used for

embedding maps on the visualisation application whereas Microsoft Azure cloud platform

will be used for hosting data on its SQL database storage and providing mobile services

to the mobile device. Also, the research will only focus on categorical, geographical and

the temporal elements of the dataset.

1.5 Structure of Thesis

The rest of the thesis is structured as follows: Chapter 2 is a literature review and

focuses on the knowledge required in understanding mobile visualisations and interaction

techniques. Chapter 3 discusses the methodology employed for the research work. The

chapter also discusses the design process carried out for the visualisation application

proposed and delves into the details of development of the application. Chapter 4 presents

the results of the evaluation conducted for the visualisation application and discusses the

relevance and meaning of the findings. Chapter 5 concludes the thesis by highlighting the

relevance of research, revising the research work’s original goals and discussing potential

future work that could be derived.

Chapter 2

Related Work

2.1 Introduction

The first research goal of this research was to review the state-of-the-art mobile visuali-

sations in the mobile platform. This chapter provides a review of such visualisations. It

gives a broader understanding of visualisations in the computing field, and then narrows

the focus to the mobile platform.

The human sense of sight accounts for more information acquisition than the rest of the

senses put together (Ware, 2013). Advances in the computing field and mobile platform,

combined with the power of visualisation and interaction techniques for mobile phones,

provides mobile users with an opportunity to access and acquire information and manipu-

late data, at a glance. In order to take advantage of this opportunity, research is required

to explore measures that can be taken to design visualisations for limitedly resourced

mobile devices. The design of a visualisation needs to be effective in providing users with

relevant information. The next section (Section 2.2) provides a broad discussion visual-

isations across all platforms in order to set context and understanding for visualisations

in the mobile platform. A discussion that is specific to the mobile platform will follow in

Section 2.3.

Data demands from various disciplines are amongst key drivers of the technological ad-

vances in the computing field. Visualisations are useful for gaining insight from the

available data, as they help users derive more knowledge from large datasets by mak-

ing complex patterns discoverable to the users (Card, Mackinlay, & Shneiderman, 1999).

12

2.2. COMPUTER VISUALISATION 13

More discussion on data types and visualisations associated with the various data types

is provided in Section 2.4.

Mobile devices are becoming ubiquitous (Mouton, Sons, & Grimstead, 2011). While there

may be other factors contributing to this ubiquity, developments in computing resources

of mobile devices feature amongst those factors. These developments together with inter-

action techniques, enable visualisations to be rendered on mobile devices (Rukzio, Broll,

Leichtenstern, & Schmidt, 2007). In spite of the advancements, mobile devices still lag

behind desktop devices due to comparatively limited capabilities (Van Tonder & Wes-

son, 2008). Visualisations are often less powerful and complex on the mobile platform

compared to the desktop platform. However, complexity does not necessarily imply effec-

tiveness for a visualisation.

An effective visualisation supports user tasks in a seamless manner. A user should be

able to focus on the intended goal during the interaction with a visualisation. In order to

achieve the goal, a visualisation should be designed such that the user focuses more on

the goal and less on the tasks. A discussion on interaction techniqeus will be given in 2.5,

followed by a discussion of user tasks commonly supported in visualisations in Section

2.6. The chapter concludes in Section 2.7, summarising the important aspects discussed

throughout the literature review.

2.2 Computer visualisation

Computer visualisation is concerned with “computer-based, interactive visual represen-

tations of data to amplify cognition” (Card et al., 1999, pp. 7). The desktop platform

has more powerful computer resources than the mobile platform. As a result, a visualisa-

tion design process targeted for the desktop environment becomes less challenging than

designing for the mobile platform (Van Tonder & Wesson, 2008). Furthermore, process-

ing of a visualisation is faster due to the processing speed delivered by the memory and

other performance enhancing factors such as advanced graphical application programming

interfaces (APIs) (Van Tonder & Wesson, 2008).

Data is a backbone to a visualisation. It is therefore important for visualisation designers

to understand data, together with various aspects surrounding it. In addition, other

important factors to consider are devices for displaying visualisations, interactions styles

on devices and users of visualisations. Considerable attention to these factors would

produce designs that achieve the desired goals which enhance acquisition of information


knowledge and aid cognition. Some studies have proposed approaches to classifying data

(Chittaro, 2006; Ware, 2013). Classes from these studies were based on the types of data

used (e.g., text and map-based) by Chittaro (2006) while Ware (2013) gives classes of

data as categorical, real-number and integer data based on wide usage of these classes of

data in the computing field. Having a clear understanding of various types of data used

assists designers and developers to create more intuitive visualisations that represent the

underlying data as closely as possible.

Interactivity in visualisations is a major factor that designers have to cater for in order to

enhance user experience. It helps a user to focus on the task at hand rather than get frus-

trated by interaction styles that require an effort for the user to learn. Ideally, a designer

should have an image of a typical user of the visualisation being developed, including a

user’s context in order to incorporate appropriate interaction techniques (Chittaro, 2006;

Dix et al., 2004). The interaction styles are supported by the type of hardware and soft-

ware available on the device. The next subsection discusses hardware and software that

is useful in visualisations.

2.2.1 Hardware and software specifications

Presentation plays an important role in a visualisation. Visualisations on the desktop

environment benefit from large screen sizes and high resolutions available in this environ-

ment (Callegaro, 2013). The powerful computing resources on the desktop platform (Yoo

& Cheon, 2006) make it arguably easier for developers to create visualisation in this en-

vironment. Designers are not confined by limitations that challenge designs in the mobile

platform.

Random access memory (RAM) plays a significant role for performance of a visualisa-

tion. Depending on how data is represented in a visualisation, RAM becomes useful in

determining how much of the representation can be cached for fast access. This aspect

improves the responsiveness of a visualisation. Desktop devices enjoy relatively unlimited

speed compared to their mobile counterparts. Processing speeds from central processing

units (CPUs) and graphics processing units (GPUs), like RAM, enhance performance

of visualisations by improving the speed at which a visualisation operates. GPUs also

improve the quality of the graphics rendered on the screens (Kutter, Shams, & Navab,

2009). These factors contribute towards the effectiveness of a visualisation, which in

turn enhances user satisfaction. Software also impacts the type of visualisations that can


run on a device. Environments such OpenGL, Direct3D, Compute Unified Device Ar-

chitecture (CUDA), Close to Metal (CTM) and GeoTools-Lite give developers access to

powerful graphics enhancing APIs that benefit visualisations (Kutter et al., 2009; Rhyne

& MacEachren, 2004).

General advances in computing areas such as grid and cloud computing and networking

(Mouton et al., 2011) provide more computing resources to less powerful workstations.

This enables more data to be accessed and processed, and in turn enabling more visuali-

sations to be developed and rendered, both locally and remotely. The resources available

on devices impact the visualisation design since designer need to match the capabilities

of the devices to the features of a visualisation. The next subsection explores factors

important for the design of visualisations in the computing field.

2.2.2 Design considerations

In human-computer interaction, the golden rule of design is ‘understanding your materi-

als’ (Dix et al., 2004, p. 193). This refers to both humans and computers. The rule can

be applied to the visualisation design process. Understanding the user needs, capabilities

and user influences can produce successful visualisations that meet users’ expectations.

Similarly, understanding the computer limitations and capabilities helps in a design of a

visualisation that also contributes towards meeting the user’s expectations.

Shneiderman (2004) discusses three goals that designs should address for successful user

interfaces. The goals can be adopted in designing visualisations as a user interface is a

predominant feature of a visualisation. The goals are described below:

1. provide the right functions that will allow users to accomplish their goals

2. offer usability and reliability to prevent frustration (from undermining fun)

3. engage users (with fun-features)

The first two goals require a visualisation designer to provide users with adequate func-

tionality to perform various user tasks related to a visualisation, e.g. providing a user

with an overview of data in a visualisation (Shneiderman, 1996, more user tasks are

discussed in Section 2.6). The functionality should also be usable and reliable and not

put mental burden on a user during interaction with a visualisation. For the last goal,


the aspect of fun could as well be brought in the visualisation domain as a way to provide

more user satisfaction. This goal is more suited to mobile visualisation since the mobile

platform is dominated by users operating in non-professional settings than the desktop

platform (Shneiderman, 2004), though mobile users can still engage in professional set-

tings. Combining goals by Shneiderman (2004) and Dix et al. (2004) gives a broader

perspective for the designer, as the designer will have an understanding of user needs,

system capabilities and user interface goals to base a visualisation on. This approach can

result in positive reception of a visualisation by users as they would have contributed to

its design.

Another study (Carr, 1999) proposes seven guidelines which can be followed in the design

of visualisations. The guidelines are described below:

1. visualisation is not always the best solution - if there are other ways of meeting user

goals, a design should explore those alternatives.

2. user tasks must be supported - refers to providing more interaction with a visu-

alisation than the basic tasks provided in Shneiderman (1996, discussed in 2.6).

This point compares with a point on interactivity in Chittaro (2006) as it addresses

interaction styles that a design should allow during the visualisation process.

3. the graphic method should depend on the data - a visualisation should represent

data in a dataset as closely as possible to the concept the data represents.

4. three dimensions are not necessarily better than two - a visualisation should use few

dimensions to represent data, but should ensure that a meaning derived from such

a visualisation is not reduced or distorted (Peng, Ward, & Rundensteiner, 2004)

5. navigation and zooming do not replace filtering - a visualisation should enable users

to not only navigate the visualisation, but also filter out data to gain more perspec-

tive of items of interest.

6. multiple views should be coordinated - if a visualisation provides multiple views

(e.g., Overview+Detail as described below), the views should be managed so that

they do not frustrate the user while switching between the views.

7. test your designs with users - testing a visualisation with users helps discover us-

ability problems. This is an important guideline in any design, as highlighted under

the guidelines by Shneiderman (2004) on usability and reliability, discussed earlier

in this section.


Figure 2.1: (a) The zoomed part gives a detailed information (Baudisch et al., 2001) (b)Fisheye view shows more detail at the centre - the focus (Ross, 2008)

Effective visualisation designs have to capture the attention of a user and deliver infor-

mation within a reasonable period of time. A user-centred approach helps to achieve this,

as it places the user at the forefront of the design process, instead of relegating the user’s

role to the periphery of the entire process. For this to occur, a communication channel

has to be established as early as possible in order to inform the design process (Chittaro,

2006; Dix et al., 2004), in an effort to understand the user.

Another advantage of engaging the user early during the design process is to help de-

signers identify usability problems associated with the initial design early on (Rhyne &

MacEachren, 2004). This helps to manage the design costs and minimise redesign costs.

Such costs include time spent on the (initial) design, financial resources incurred and

human labour deployed to the design process. Through the established communication,

the developer can learn what the user’s needs are and the user can in turn be aware of

what the developer can and cannot offer. This approach leads to a better understanding

of both players involved and hence a more successful visualisation.

Approaches specific to map-based visualisations have been discussed in various stud-

ies (Baudisch et al., 2001; Chittaro, 2006). These are Overview+Detail (O+D) and

Focus+Context (F+C). They are meant to overcome the presentation issue on desktop

visualisations. O+D has one part of the screen displaying the overview of the entire visu-

alisation while another part of screen displays a detailed part of the visualisation. F+C

2.3. VISUALISATIONS IN THE MOBILE ENVIRONMENT 18

provides one view that shows both the context or overview together with content which

is the detailed part of visualisation (Chittaro, 2006). Examples of O+D and F+C are

shown in Figure 2.1 (a) and (b) respectively. The next subsection discusses categories of

visualisations in the computing field.

2.2.3 Categories of visualisations

According to Card et al. (1999), computer visualisation in the early days focused on sci-

entific computing with the aim of helping scientists work with large amounts of scientific

data. Scientific visualisation therefore tended to be based on the nature of data scientists

worked with, which mostly constituted physical objects (Card et al., 1999). Visualising

data that was not based on physical objects but rather on abstract concepts (such as

financial data) became apparent. Visualisation from the abstract concepts was coined

information visualisation (Card et al., 1999). Other forms of visualisations exist. These

include volume, flow, surface and geographical visualisation (Card et al., 1999; Rhyne

& MacEachren, 2004). These visualisations either work with physical data or abstract

concepts and can therefore be categorised as information or scientific visualisation. Vol-

ume visualisation deals with representation of 3D volumes and interaction techniques for

working with those volumes. Flow visualisation deals works on visualising flows (of liq-

uids or gases). Visualisation of data from geographical or location based disciplines is

called geographical visualisation. Categorisation by Card et al. (1999) essentially gives

two categories of visualisation.

The next section discusses visualisations in the mobile platform, followed by hardware and

software specifications, design considerations that affect design in mobile visualisations

and categories of visualisation in the mobile platform.

2.3 Visualisations in the mobile environment

The limited resources in mobile devices pose a challenge for visualisations, especially

in comparison to the desktop devices. The challenge lies developing new visualisation

techniques, since the traditional desktop visualisations cannot simply be transferred to

the mobile platform given the limited resources such as small screen sizes and proces-

sors (Chittaro, 2006; Van Tonder & Wesson, 2008). Despite this challenge, technological

advances in the computing field offer some opportunities to the mobile platform. For


example, wireless networking technologies provide a platform for development of cloud-

based applications which are unique to the mobile platform (e.g., navigation services)

(Chittaro, 2006; Mouton et al., 2011).

The mobility factor also adds a new dimension of complexity. Mobile application devel-

opers have to consider factors such as multitasking, short attention spans and changing

lighting conditions on a device screen (other factors are discussed in other sections, e.g.,

interactions styles are discussed in Section 2.5). These factors give rise for the need to

design applications that give as much information as possible, and in arguably a shorter

time than on a desktop device (Chittaro, 2006). Visualisations could achieve this short

time requirement since they enhance acquisition of more information in a relatively shorter

time than if presented otherwise (Card et al., 1999). The next subsection discusses hard-

ware and software specifications that are useful in visualisations, followed by factors that

affect the design of mobile visualisation and a description of categories of visualisations

that are available in the mobile platform.

2.3.1 Hardware and software specifications

The ability of a mobile device to render visualisations depends on the hardware and soft-

ware available on the device. The hardware and software on a mobile device consequently

impact the design of a visualisation. The design of visualisations on the mobile platform

therefore becomes a challenge due to limited resources (Van Tonder & Wesson, 2008). Be-

low is a description of hardware and software that play various significant roles in mobile

visualisation.

1. Screen - this is perhaps the most important resource for a visualisation since it affects

the layout of visualisation (Chittaro, 2006). Screen size determines the amount of

information a visualisation can display, while resolution specifies the quality of a

graphics used for a visualisation. Higher resolution such as retina display on Apple

devices offer higher quality displays (Apple, 2013). Touchscreens are a dominant

screen type in mobile devices.

2. Processor and RAM - these are responsible for performance of a visualisation since

they determine how much and how fast data can be cached and accessed during a

visualisation process.


3. Connectivity - mobile phones, especially smartphones, typically have several con-

nectivity types such as WI-FI, mobile telecommunication technologies (e.g., 2G,

3G, 4G), Bluetooth, near field communication (NFC) (Samsung, 2013). These can

be used in connecting with other devices that provide data for visualisation, or

cloud-based visualisations (Mouton et al., 2011). Various connectivity capacities

determine the speed of a visualisation process and are key especially if cloud-based

visualisations are involved.

4. Storage - this is only useful to the extent a user needs to perform such tasks as

extracting parts of data from a visualisation on a mobile phone.

5. Input/output devices - examples include microphones and speakers, which provide

new ways of interaction, such as speech recognition (e.g. Siri by Apple) and many

types of sensors (e.g. proximity sensors, GPS, temperature and humidity sensor)

available in smart phones like Apple iPhone 5s and Samsung Galaxy S4 (Apple, 2013;

Samsung, 2013). The io devices impact the design of visualisations as designers need

to know which mobile devices to target in order to exploit as many io devices as

possible and support as many interactions as possible on mobile devices (Wang et

al., 2013).

From the software perspective, libraries and APIs exist that enhance visualisations. Ex-

amples include Google Maps API, GeoTools-Lite and OpenGL ES (Rhyne & MacEachren,

2004; Van Tonder & Wesson, 2008; Wang et al., 2013). These tools help in designing vi-

sualisations that optimise hardware resources, hence improve the overall performance of

visualisation. The knowledge and understanding of various hardware and software avail-

able on mobile devices can help in the design of new applications that take advantage of

some of the unique features (e.g., GPS) of mobile devices. The next subsection explores

factors that affect visualisation designs in the mobile platform.

2.3.2 Design considerations

In order to address some of the challenges and take advantage of new opportunities high-

lighted in the previous sections, design approaches have been proposed by several studies

(Chittaro, 2006; Fling, 2009). Fling (2009) gives guidelines that are general for mobile de-

sign and development, but could be adapted for the mobile visualisation. Chittaro (2006)

gives an approach that is specific to mobile visualisation. Descriptions of the approaches

are given in the following paragraphs.


Figure 2.2: (a) Halo with arcs referencing off-screen places (Baudisch & Rosenholtz, 2003).(b) ZoneZoom segments a screen into nine views (Robbins et al., 2004).

An approach by Chittaro (2006) proposes six area of focus when designing for mobile

visualisation. These are:

1. mapping - designers should be able to graphically emphasise important aspects

within the visualised data such that the user will effortlessly pick important patterns

and relationships within the visualisation.

2. selection - the limited screen requires careful attention. Designers should be able

to pick relevant data and discard undesired data. Selection should not only be

determined by the designer, the user should also be able to filter for relevant data

within the entire dataset, in order to suit their own needs (Fry, 2009).

3. presentation - this is a key area for visualisation as it is a final destination of informa-

tion to the user in the visualisation. Approaches such as O+D and F+C (discussed

in Subsection 2.2.2) often fail due to limited screen sizes (Chittaro, 2006). Different

approaches specific to map-based visualisations have been proposed, namely i) vi-

sual references to off-screen points of interest, e.g. in Halo (Baudisch & Rosenholtz,

2003) and City Lights (Zellweger, Mackinlay, Good, Stefik, & Baudisch, 2003). Fig-

ure 2.2 (a) shows a screenshot of Halo. The arcs along the edge of the screen direct

the user to the off-screen points of interest in a map. However, these approaches

tend to introduce some clutter of their own (Van Tonder & Wesson, 2008) and ii) in-

tuitive ways of switching among the visualisation parts, e.g. in ZoneZoom (Robbins

et al., 2004). ZoneZoom splits the view into nine sub-views which are mapped to


number keypads on a phone. Figure 2.2 (b) ZoneZoom view of a map. To zoom in

to an area of interest, a user selects a number corresponding to the area of interest.

4. interactivity - this concentrates on how users interact with the visualisation. A

visualisation should allow the user to interact with the dataset as much as possible

as they explore the information for patterns.

5. human factors - it is crucial to understand target users of a visualisation - their

literary and computer skills and cognitive abilities. The design should provide a

visualisation that is highly usable even amongst a widespread potential user pop-

ulation in the mobile platform (Dix et al., 2004; Paelke, Reimann, & Rosenbach,

2003).

6. evaluation - the effectiveness of the visualisation should be determined. It is impor-

tant to test the usability of a visualisation, based on what the user’s experience is

during the interaction (Dix et al., 2004).

Fling (2009) has seven areas for design focus for mobile platform, could be adapted for

the design of visualisations. These are:

1. context - which is defined as:

� the general understanding of the task a user is performing

� the environment in which the task is performed:

– physical context - the physical location of the user

– modal context - the psychological state of the user

A comparable study (Pombinho, Carmo, & Afonso, 2009) proposes five contexts

to explore in visualisation, these are: personal, environment, temporal, social and

spatial contexts. The first two are similar to the modal and physical contexts in

Fling (2009). The last three offer additional context, and relate to a time when a

user performs a task, other people around while a task is performed and orientation

respectively. Some of the contexts tend to be neglected in many mobile applications,

focusing interaction only between the user, the mobile device and services (Rukzio

et al., 2007). All these allude to complexity involved in designing for mobility.

2. message - what the visualisation needs to tell a user.

3. look and feel - how the interface is designed and how will users interact with it.


4. colour - colour depths, posterisation, human perception, psychology of color and

colour palettes should all be considered for an effective design.

5. layout - is the visualisation laid out such that a user will interact with it in a

simple way, without mental overload. Other researchers present similar suggestions

regarding presentation. Owing to small screens, clutter is a serious challenge in

mobile devices compared with the desktop devices. It reduces the user’s ability to

perform a task effectively (Pombinho et al., 2009; Rosenholtz, Li, Mansfield, & Jin,

2005). As such, a considerable effort is required to minimize clutter when designing

a visualisation.

6. typography - how a design addresses fonts, size, styles and readability and clutter

is managed.

7. graphics - how a design aids a user’s cognition with images - iconography.

Although Fling (2009) makes a distinction between the last four points, the points address

the same issues related to layout and so can be put as one point - look and feel. The

first three points can then be adopted for a visualisation design. Some aspects from Fling

(2009) and Chittaro (2006) are comparable. For example, the message in Fling (2009)

addresses some of the aspects of a visualisation similar to those addressed in presentation

and mapping from Chittaro (2006). Human factors and interactivity in Chittaro (2006)

address factors such as interaction styles available for a visualisation and these factors are

covered by the context in Fling (2009).

There are other factors that affect the design of the visualisation. A wide variety of mobile

devices that exist in the market is another challenge to designers of visualisations. Frag-

mentation features significantly, since designers have to worry about providing effective

visualisations across many devices in a way that does not distort meaning from one device

screen size, resolution and OS to another (Akbari et al., 2014; Callegaro, 2013).

Connectivity is also worth considering during the design of a visualisation. The cost

(of short life batteries of mobile devices, telecommunication network charges and low-

bandwidth connectivity speed) associated with rendering visualisation on a mobile phone

require a structured approach to design. The challenge posed by low band-width con-

nections can be overcome by designing visualisations that minimise operations on mobile

devices, leaving presentation as the only task done on the mobile device.

The next subsection discusses categories of visualisation that exist for the mobile platform.

2.4. VISUALISATION DATA TYPES 24

2.3.3 Categories of visualisation

A discussion on categories of visualisations was provided in Section 2.2.3, giving cate-

gories as scientific and information visualisations. This section discusses categories of

visualisation that were found in related work to be more applicable to the mobile context

as proposed by Chittaro (2006). Five categories have been proposed for mobile visual-

isation based on the following data types: text, abstract, maps, physical objects, and

pictures data. The categories can also be put scientific or information visualisation. For

example, visualisations of physical objects is scientific visualisation according to a classi-

fication given in Section 2.2.3. Depending on the context, some visualisation categories

will be more popular than others. For example, information visualisation is arguably

more popular in mobile platform than scientific visualisation due a wide population po-

tentially dominated by non-professionals (Shneiderman, 2004). This is due to the ease of

access and use by mobile devices without a need for formal training as it is a normal case

with desktop computing (Chittaro, 2006; Shneiderman, 2004). The next section discusses

data types that are typically used in visualisations, resulting in a broader domain for

visualisations than provided by categorisation from this section and Section 2.2.3.

2.4 Visualisation data types

Data is key to visualisation. Many disciplines generate and use data from their own insti-

tutions for internal and external use. A discussion in previous sections (Section 2.3 and

3.3) essentially narrows visualisations down to two categories - scientific and information

visualisations. In this section, visualisations are classified according to data types. This

gives a broader domain of visualisations than just the scientific and information visualisa-

tions. Data types commonly used in visualisations were investigated from several studies

(Chittaro, 2006; Shneiderman, 1996; Ware, 2013). Ware (2013) classifies data according

to three types, namely categorical, integer and real-number data. Categorical data as-

signs labels to data within datasets and cannot be ordered in a sequence. For example,

Figure 2.3 (a) shows a screenshot of news24 elections app described later in this section,

the different political votes represent categorical data. Categorical data type is one of two

types which will be used in this research. Integer data involves data that can be ranked

in a sequence. For example, ranking the political parties according to the total votes they

got from the elections falls under this data type, as shown in Figure 2.3 (a). Real-number

data is data associated with intervals and ratios. For instance, the distances between the

different points of interest in Figure 2.2 (a) fall under the real-number data type.


Chittaro (2006) gives five classes of data which visualisation targets on mobile devices.

These are pictures, text, physical objects, abstract data and maps. Data represented by

visualisations under this classification is in the form of pictures, text, physical objects,

abstract concepts and maps respectively. Map-based data type is the other type under

the focus for this research.

Shneiderman (1996) proposes a taxonomy based on seven data types, namely: 1-dimensional

(1D), 2-dimensional (2D), 3-dimensional (3D) and multi-dimensional, temporal data, tree

data and network data. 1D type represents data that has one variable in a database, e.g.

list of honours students for a particular year. 2D type represents data with two variables,

visualisations include geographic data or surface areas, with x and y coordinates that

represent a point on a surface. Figure 2.2 shows examples visualisations of 2D data type.

3D works with data that has three values or variables in a representation, e.g., an object

has three x, y and z coordinates that represent a point in space. Figure 2.3 (b) shows

an example of a visualisation for 3D data. Multidimensional types represent data stored

in statistical and relational databases, usually with more than three variables. Visuali-

sations for this data type are usually expressed as scatter plots and parallel coordinates.

Figure 2.4 shows parallel coordinates visualisation for automobile data as an example of

multidimensional data type. Temporal type includes any of the above data types, plus an

added time component. This time is used to explore behaviour of data as time changes.

Figure 2.5 shows a temporal data visualisation for selected technology stock between the

years 2000 and 2010. Tree data can be displayed with hierarchical relationships that show

links between items in a dataset. Each item has one link above it - to a parent item, and

can have other links below - to child items. An exception is the parent item, which has no

parent link. A directory structure in a computer filing system is an example of such data

type. An example of a tree visualisation is shown in Figure 2.6. Network type represents

links between groups of items in a dataset. Nodes in this type cannot be represented in

a tree-like format. An example of visualisation is shown in Figure 2.7.

Data types provided Shneiderman (1996) encompasses classifications given by Chittaro

(2006) and Ware (2013). For instance, all examples given in the classification by Ware

(2013) can be represented in any of the type given by Shneiderman (1996). For a classifica-

tion provided by Chittaro (2006), pictures, maps and physical objects can be represented

in 2D or 3D, whereas text and abstract data can be represented in 1D.

The data types to be visualised in this research are categorical and map-based data types.

As such, a detailed investigation was conducted for an application that focused visualisa-

tions on these two data types. News24 Elections mobile application for visualising South


Figure 2.3: (a) News24 elections app displaying 2014 results by provinces (map-based)and total number of votes (integer data) political parties got (categorical data) (News24,2014). (b) 3D land surface visualisation of wind vector profiles (Wang et al., 2013)

African 2014 national election results and related news was explored. The application

makes use of categorical data (e.g. different political parties participating in the elec-

tions) and map-based data (e.g., provinces, municipalities, etc.) to provide visualisations.

Colours on various map divisions indicate the political party that had the most votes.

For instance, Figure 2.3(a) shows national and international results. The international

result shows an overall blue colour, while the national results show an overall green colour,

except in one province. The underlying data used in this application is based on a large

dataset of over 25 million registered voters. This application closely relates to the research

topic, which focuses on visualisations of large datasets of map-based and categorical data

types. Interaction styles such as pinching and tapping are provided by the visualisation

2.5. INTERACTION TECHNIQUES FOR MOBILE DEVICES 27

Figure 2.4: Parallel coordinates visualisation for automobile data (Heer et al., 2010).

to enable a user zoom in and view more details associated with geographical divisions

of interest to the user. The next section discusses interaction techniques available in the

mobile platform and how they support visualisation in the mobile devices.

2.5 Interaction techniques for mobile devices

Interaction techniques provide a means for a user to actively engage in a visualisation

process. A user is able to investigate, evaluate and drill down more on the data being

visualised using some graphical user interface elements such as sliders and buttons or

hardware buttons on mobile devices. The fragmentation problem from having many

devices by different manufacturers brings a wide variety of interaction techniques which are

an advantage and a disadvantage at the same time (Akbari et al., 2014). The advantages

lie with users, since there is a wide variety of devices to choose from, whereas for the

designer of a visualisation, incorporating a wide variety of devices is a challenging task

(Akbari et al., 2014). These techniques can be exploited to enhance user experience.

As highlighted in Section 2.3.1, interaction techniques depend on the hardware and soft-

ware available on a mobile phone. Often a combination of hardware resources achieves

a single interaction technique, so interactions described under one resource are not ex-

clusively realised by that resource. For instance, sensors detect device orientation when

2.5. INTERACTION TECHNIQUES FOR MOBILE DEVICES 28

Figure 2.5: Visualisation of technology stocks representing temporal data (Heer et al.,2010).

a user tilts a screen, and a screen display changes from horizontal to vertical layout or

vice versa. Touchscreens support interaction techniques such as touch (e.g. swipe, slide,

pinch, tap, multitap, etc. (Rukzio et al., 2007)). Some touchscreens also adapt to user

context such as adjusting to lighting conditions.

Sensors enable techniques such as orientation (e.g. tilting and shaking) and motion de-

tection (e.g. eye blink rate sensors in Google Glass) (Ishimaru et al., 2014; Rukzio et al.,

2007). Some of the interactions allow for applications that are unique to this platform.

For example, GPS allow mobile phones to interact with their geographical location, and

provide services such as navigation. Applications that take advantage of these interac-

tions techniques have been explored, such as TiltText (Wigdor & Balakrishnan, 2003)

and TiltType (Partridge et al., 2002). For example, TiltText (Wigdor & Balakrishnan,

2003) uses an accelerometer and a tilt sensor to allow a user to select an alphanumeric

option while texting a message on a mobile phone. Figure 2.8(a) shows how a user tilts

a phone to input one of four options associated with a keypad.

Cameras enable users to use mobile devices as pointing device to other application (Bal-

lagas, Borchers, Rohs, & Sheridan, 2006). An example is shown in Figure 2.8(b), where

a camera is used in augmented visualisation to give a user a different visualisation from

the one on a wall display (Soros et al., 2011). Designing visualisations that exploit these

techniques will enable a user of visualisation to have many options while interacting with

2.6. USER TASKS IN VISUALISATION 29

Figure 2.6: Tree visualisation (Heer et al., 2010).

Figure 2.7: Network visualisation (Heer et al., 2010).

the visualisation. The interaction techniques discussed in this section typically support

tasks a user undertakes in order to achieve thier goal for using a visualisation. The next

section discusses user tasks commonly used in visualisations.

2.6 User tasks in visualisation

A user engages with a visualisation because they have a goal of gaining insight into the

data being visualised. The goal aims to help a user make discovery of patterns within the

dataset, make decisions or find explanations within the data (Card et al., 1999). To achieve

the goal, often a series of tasks are undertaken. A visualisation should be designed such

that a user’s attention is focused more on the goal and less on the tasks involved in achiev-

ing the goal. Seven general visualisation user tasks that abstract the visualisation design

2.6. USER TASKS IN VISUALISATION 30

Figure 2.8: (a) User tilts a phone to select letters as he types text (Partridge et al., 2002).(b) Augmented visualisation (Soros et al., 2011)

process from any specific visualisation have been developed by Shneiderman (1996)(Carr,

1999). The seven tasks are described below according to Shneiderman (1996):

1. overview - a visualisation should allow a user to have an overview of the entire

dataset, usually through zooming to the outermost level.

2. zoom - a user should be able to focus in detail on parts of interest in a visualisation.

3. filter - this complements zoom in that a user should be able to filter out items that

are not of interest.

4. details-on-demand - a user should be able to get the details of an item or a group

of times of interest when needed by selecting the item or the group.

5. relate - a visualisation should provide a way to explore relationships among items

in a dataset.

6. history - this ensures that in a common but undesirable event when a user performed

a certain action or got unexpected results from that action, they will be able to undo

it.

7. extract - this allows for querying of data within datasets. Users should be able to

extract parts of visualisation data if they so require.

A summary of literature review is provided in next section, highlighting important issues

discussed throughout the literature review.

2.7. SUMMARY 31

2.7 Summary

This chapter explored literature related to visualisation and interaction techniques in the

desktop and mobile environments. Specifically, the mobile platform as the focus of this

research topic was investigated in detail. In addition, hardware and software that is use-

ful for improving performance and quality of visualisations on the mobile platform was

investigated. Design considerations on both platforms were discussed, looking into the

guidelines that have been developed and are used in visualisations. Visualisation cate-

gories were also investigated for both environments. Emphasis was placed on managing

clutter, especially in small displays of mobile devices. Data types common in visualisa-

tions were explored, highlighting categorical and map-based data types as key types to be

visualised for this research. Examples were drawn from various applications to visualisa-

tions. Interactivity as an important factor in visualisation was discussed, highlighting its

importance in assisting a user to achieve their tasks. The literature review concluded with

a discussion on the user tasks commonly used in visualisation, emphasising the importance

of having those fundamental tasks in all visualisations.

The findings from this chapter are used to inform the design and implementation of a

visualisation application discussed in the next chapter. The methodology employed during

this research project will be discussed, together with the design and implementation of

the visualisation prototype in Chapter 3.

Chapter 3

Methodology, Design and

Implementation

3.1 Introduction

This chapter discusses the methodology deployed throughout this research. The review

of literature from the previous chapter was used to influence design, implementation and

evaluation of the prototype described.

3.2 Spiral Methodology

The methodology that was followed to achieve the research goals is the spiral model. The

model, developed by Boehm (1988), goes through more than one iteration or spiral in

each of the four phases described below. Figure 3.1 shows a diagram of a spiral model.

The four main phases of the model are:

� identification - this phase identifies the system requirements or the objectives and

constraints of the project. Section 3.3.1 will discuss this phase of the project.

� design - a prototype of a desired product was developed in this phase. A visualisation

prototype was designed, based on the findings from the previous phase. A broad

discussion will be given in Section 3.3.2 of this chapter.

32

3.3. ITERATION 1 33

� development or build - this phase deals with the implementation of a prototype.

Implementation starts after the design is completed in order to realise the design.

This phase will be discussed in Section 3.3.3 of this chapter.

� analyse - the prototype is evaluated against risks it may face before it is released to

the users. A detailed discussion on how the evaluation was set-up and carried out

is given in Section 3.3.4

Figure 3.1: The spiral model (Sayyed, 2012)

The iterative nature of the model is appropriate since it provide flexibility within this re-

search as development of the visualisation prototype was produced early in the application

life cycle and then incrementally refined later on - incorporating emerging needs or needs

that were not identified during at the beginning of the research work. Two iterations were

followed in this research project as described in Sections 3.3 and 3.4.

3.3 Iteration 1

This iteration focused on the literature review, design and development of the mock-up

prototype and an evaluation of the mockup prototype using a pilot study with two expert

evaluators.

3.3. ITERATION 1 34

3.3.1 Identify

As previously discussed in Chapter 2, visualisations such as F+C and O+D (Baudisch

et al., 2001) were targeted for the desktop platform and cannot be directly ported to

mobile phones due to the small screens on the devices. Visualisations that were targeted

for the mobile platform, for example Halo (Baudisch & Rosenholtz, 2003) and ZoneZoom

(Baudisch & Rosenholtz, 2003), were either ineffective or introduced clutter on the screens.

One of the limitations on the mobile platform, and in particular to visualisations, is the

small screen on the mobile devices. Paying attention to what goes into the screen therefore

needs to be carefully considered in order to avoid the same problems faced by previously

mentioned mobile visualisations. Other challenges posed by the limited resources will also

be considered as the prototype is designed and implemented.

Interaction techniques have been shown to be useful in supporting users in performing

tasks as they interact with visualisations. It is therefore important to identify interaction

techniques that will support the proposed visualisation. As the target mobile phones are

smartphones, most (if not all) have touchscreens. Touchscreens support a variety of ges-

tures (e.g. swiping, tapping and pinching) as a means of interacting with the smartphone.

User tasks, as described in Section 2.6, can easily be supported if the right interaction

styles are chosen within a visualisation. The interface for the proposed visualisation ap-

plication was influenced by the News24 Elections mobile application mentioned in Section

2.4 as it uses geographical and categorical data types for visualising the election results.

The proposed visualisation for this research project uses geographical and categorical data

types for visualising water quality results. The geographical part is made up of twenty-

three suburbs (see Listing B.2) around Grahamstown, while the categorical part is made

up of four categories (see Listing B.3) taken from the MobiSAM website water quality

polls. The visualisation design is described in the next section.

3.3.2 Design

An ideal prototype would provide a user with at least three features that enable users to

visualise and interact with data with respect to the geographical and categorical data types

as they (might) require separate techniques for visualisation and interaction. Providing

users with these three functionalities would ensure that users accomplish their goal -

which is to visualise data. This is in line with two of three goals for designing interfaces

by Shneiderman (2004) discussed in Section 2.2.2, namely, to provide the right functions

3.3. ITERATION 1 35

Table 3.1: Sample responses from Vukani suburbCategory Responses % ResponsesSmells offensive 7424 29.1brownish 7101 27.9cloudy 6294 24.7Smells alkaline 4634 18.2

that allow the users to accomplish their goals and to offer usability and reliability to prevent

frustration to users. The most important feature for this project is the ability to provide

the user with a map on which data overlays and other features would be superimposed.

As presentation has been shown to be one of the key design considerations by Chittaro

(2006) (from Section 2.3.2), it is important to provide the necessary functionality without

cluttering the interface with a lot of widgets. It should therefore be enough to provide only

the map from which users will interact with data by tapping suburbs as they interrogate

the visualisation. Tapping the interface allows a user to engage with a visualisation

interactively and is a key design consideration for visualisations.

Displaying the boundaries of the individual suburbs is also crucial so that users will be

able to identify them. Displaying the boundaries also forms part of the presentation

aspect mentioned above, while identifying and interacting with them through tapping

addresses the interactivity design consideration. Each of the four categories is represented

by a specific colour within the visualisation. Within a suburb, a colour that represents

a category that has the highest number of responses colours the suburb polygon. To

illustrate, suppose a suburb Vukani has received responses classified according to Table

3.1. Out of all the responses, the highest number of responses from that suburb is in

the category of water that ‘smells offensive’ and so the polygon that represents Vukani

would be coloured with a colour that represents this category. The other aspect related

to colour is the intensity of the colour. The design is such that the intensity increases

with the percentage it represents, such that the smaller percentages are more transparent

than larger percentages.

A bar chart is placed at the top of the visualisation. This bar always displays the same

overall results of the entire dataset to provide an overview of the overall situation as one

of the user tasks by Shneiderman (1996) described in Section 2.4. The chart displays

the top two categories and sums the rest (if they exist) as ‘Others’ and displays their

corresponding percentages and colours.

Interactivity has been shown to be one of the key success factors in designing visualisations

(Chittaro, 2006). To provide interactivity in the visualisation, the suburbs should be

3.3. ITERATION 1 36

interrogated (through tapping them) to display their information both on the map and in

textual format. Another interaction with the map is the zooming feature. Users can zoom

in and out if they wish so that they can closely inspect their regions of interest. Other

interaction techniques include scrolling both on the map and on the list items displaying

details related to the suburbs or categories.

Figure 3.2: (a) Screen for displaying results for all suburbs b) screen displaying resultsfor one suburb

3.3.3 Develop

The Balsamiq Mockups application was used for the design of the prototype. The appli-

cation is a graphical user interface builder that allows designers to build mock-ups using

widgets in a WYSIWYG editor. The design of the visualisation was developed using the

Balsamiq Mockups as it provided the required widgets for the visualisation application.

Figure 3.21 shows a prototype of the visualisation. The prototype was developed as it

makes the design process faster. Designing using a mock-up application is faster than

coding. Since little time is spent on the design, it becomes easy to change the design

based on feedback. The implementation of the prototype will be discussed in Section

3.4.3. The next section discusses the user study undertaken for the mock-up prototype.

1The suburb layout used for the design does not match the Makana suburbs but was sufficient for thedesign. The actual suburb layout was used at the time of coding.

3.4. ITERATION 2 37

3.3.4 Analyse

Following the development of the mock-up prototype, an expert evaluation was undertaken

with my supervisors and more features and functionality incorporated into the mock-up.

The mock-up was found to only focus on the suburbs for visualisation, that is, suburbs

were visualised and their details displayed on the listview. The mock-up did not provide

a distinct technique for visualising categorical data. As such, a suggestion was made to

incorporate this feature. Another addition was the inclusion of the temporal component

in the dataset. The button (see Figure 3.2) at the bottom of the screen was deemed

to be unnecessary, instead, a scrollable listview was recommended. These changes and

additions will be implemented in the second iteration.

3.4 Iteration 2

This iteration describes work based on feedback from the last phase of the fist iteration of

the spiral model. The visualisation application was then designed and developed using the

development tools. This iteration will conclude with a discussion on the user evaluation

of the visualisation application.

3.4.1 Identify

The feedback from the last iteration mandated some additional features. A new feature

that allows different visualisations for the suburbs and categories was identified. The

initial mock-up only provided a means for visualising data by suburb, so an additional

means for visualising data by categories would be designed. Incorporating the temporal

element was another feature identified as missing from the evaluation of the mock-up

prototype.

3.4.2 Design

To implement the feature for visualising data by different data types, two tabs were

provided - one for categories and the other for suburbs. These would allow for separate

interactions for the two data types. For example, visualisation in the categories tab would

be achieved by tapping a suburb on the map displayed, while in the categories tab a user

3.4. ITERATION 2 38

would have to select a category from a drop-down menu to visualise the details. Another

feature added in this phase was filtering of the dataset using the temporal component in

the dataset. The filtering would be done using a slider widget. The slider was designed

to provide a window of up to ten previous days for visualising data.

3.4.3 Develop

A visualisation prototype will then be developed following the design. The implementa-

tion of the prototype will depend on the availability of resources such as the technical

expertise and skills of the researcher and development tools at the disposal. As mentioned

in Subsection 2.3.2, development of mobile visualisations is constrained by the limited re-

sources. The development will consider the limitedness of mobile resources. One way

to achieve this will be the reliance on APIs already developed for the platform to build

the desired functionalities. Following these suggestions, a visualisation prototype was

ready to be developed. The next section discusses the implementation of the visualisation

mock-up.

Development Tools

Eclipse ADT

As Android was chosen as a target platform, the Eclipse Android Developer Tools2 (ADT)

Bundle was chosen to assist with the development task. Java was therefore the primary

programming language used for coding the visualisation. The first reason is that the ADT

is freely available. The other reason is that it includes an Android software development

kit (SDK) integrated with the Eclipse integrated development environment (IDE) that

provides tools for developers to develop, test and debug Android applications from the API

libraries in the IDE (Android Developers, 2014). All coding was done using the Eclipse

ADT through its extensive suite of libraries. As interactivity is key to the visualisation

prototype, the widgets described in the designed prototype were meant to be as interactive

as possible. For example, the slider described in the design was implemented using a slider

widget. Another important interactivity is provided through the map as discussed in the

section below.

2Available at http://developer.android.com/tools/sdk/eclipse-adt.html

3.4. ITERATION 2 39

Google Maps Android API v2

As mentioned in Section 3.3.2, a map is the most important feature for this project because

of the geographic data in the dataset. Google Maps (in comparison with competitors such

as deCarta, Open Street Maps, Apple Maps and Bing Maps) was chosen as most ideal to

incorporate a map within the visualisation prototype due to its integration within the An-

droid platform. The API is also freely available. Google Maps Android API v2 provides

developers with easy integrations of Google Maps data to their apps. Google Maps An-

droid API v2 provides designers with the capability to embed maps in their applications

and add other features like markers, polygons, polylines, etc. that are typically available

on Google Maps through the API (Google Developers, 2014). The API also supports the

interactivity designed for in the design section of this chapter. The specific map events

supported by the API include clicking (or touching on touchscreens) a geographical point

on a map, zooming in and out, pinching as a zoom gesture on touchscreens and scrolling

or moving the map around to reveal other parts that may be hidden from the current

display. Providing these would enhance users’ experience with the visualisation as many

users are already familiar with using Google Maps. OpenGL is a platform powerful for

graphics useful to developers (Rhyne & MacEachren, 2004) and the API uses it (OpenGL

ES version 2) for displaying the maps (Google Developers, 2014). The API also handles

automatic tasks like connecting to Google Maps server and responding to map events

(Google Developers, 2014). In order for any application to communicate with the API,

a Google Maps Android API key is required and is freely available from Google APIs

console. Google Play services SDK is also required to use the API. The data used in the

visualisation for colouring the suburbs is hosted in Microsoft’s cloud platform - Microsoft

Azure. Microsoft Azure is discussed in the next section.

Microsoft Azure

To host the data used for the visualisation application, Microsoft Azure was used for

its online SQL database storage. A database was created and three tables added for

the project. The three tables created store all the data used in the visualisation. The

tables are depicted in Figure 3.3. At the time of writing this thesis, twenty-three suburbs

were stored in the Suburbs table and four response types/categories were stored in the

ResponseIdsWaterQuality table. The Responses table had over 420,000 records for use in

the visualisation. This data was generated for the research project.

The other important use of Azure is the mobile services it provides to developers of

3.4. ITERATION 2 40

Figure 3.3: (a) Relationships in the visualisation application tables

mobile applications. Mobile services provide a backend for mobile applications for various

purposes like push notifications, authentication of users, storing data on the cloud and

adding custom business logic to the application (Mirosoft Azure, 2014). Windows Azure

allows developers the flexibility of choosing between the .NET and JavaScript backend for

creating logic to the mobile service. The JavaScript backend was used for this research as

the researcher was more familiar with it than the .NET framework. SQL was also used to

process requests sent to the mobile service from the application into desired output. The

mobile service was used primarily for its ability to allow custom APIs created. Custom

APIs provide logic not supported by the typical CRUD3 table operations (Mirosoft Azure,

2014). The only HTTP method used was a POST as there was no need for other methods

since users should not (need to) modify (e.g. update, insert or delete) any data using

other methods such PATCH, PUT or DELETE. The APIs created for the prototype are

listed in Appendix ??. The two APIs are responsible for processing requests and sending

results back to the mobile application for visualisation. Microsoft Azure supports API

calls for raw JSON. As such, requests sent to the mobile service are JSon objects passed

to the API call as object body created in the application. JSON was chosen for data

exchange since it is easier to use that XML and it also reduces the need to create objects

within the programming environments in which they are used. A JSON body used in the

visualisation contains properties that need to be passed to the SQL statements used in

querying the database. The properties include the suburb id (1 - 23) if a user tapped on

a map to get details for a suburb or a category id (1 - 4) if a user interrogates details

related to one of four categories. Another property added to the JSON object is the date

(if a user has chosen to filter the dataset using the slider).

Microsoft Azure provides some level of security of the mobile service in order to protect

the data from vulnerabilities that may not be foreseen during development. For example,

permissions for the HTTP methods used in a custom API can be set for various user

3create, read, update and delete operations typically available in many database management systems

3.4. ITERATION 2 41

types such as ‘everyone’, ‘anyone with the application key’, ‘only authenticated user’ and

‘only administrators’. For this project, the level was set to anyone with the application

key for the custom APIs created as they only expose the HTTP POST method that does

not modify data within the accessed tables. The research cloud-based mobile service is

available at http://mobisam.azure-mobile.net/.

Other resources

The coordinates for the vertices of the suburb boundaries were downloaded from the

MobiSAM website. One important method that is not built-in to Google Maps Android

API is determining if a clicked point is inside a polygon and was sourced from the developer

site. The method was adapted to match the Latitude/Longitude point format used by

Google Maps Android API as the original method used the Point class of the Java library.

The Visualisation application architecture

The system architecture of the visualisation system is depicted in Figure 3.4. The figure

shows how all the components are communicated via the Internet. This therefore implies

that the prototype relies on constant availability of the Internet connection in order to

function, hence some potential hindrance for access to a wider user base. The tools

and resources described in the previous sections have used to produce the visualisation

application depicted in Figure 3.5. This will be discussed further in the next section.

The Visualisation Application

The final visualisation application includes:

1. two tabs to allow a user to visualise details either by suburb or category.

2. a bar chart that displays the overall results for the suburbs;

3. a map that displays Grahamstown, together with the suburbs boundaries;

4. a time slider that allows a user to select a window of time from which to visualise

the dataset; and

5. a listview that displays information related to suburbs or categories

3.4. ITERATION 2 42

Figure 3.4: System architecture for the entire visualisation application

Screenshots for the visualisation are shown below in Figure 3.5 as an example. Figure 3.5

(a) shows the overall view of the dataset. The details in the listview widget show total

responses for different categories together with their corresponding percentages. Figure

3.5 (b) shows details of changing in listview for filtering the dataset for the past six

days. As described in the design section, this updates the details displayed such that the

data visualised is for the last six days. Figure 3.6 (c) shows a screen in which details

for categories are visualised. The concept is similar to that of visualising details by

suburbs, where a user selects a category of interest and then the suburbs are displayed

with respect to that particular category. The details (suburb names) in the listview are

scrollable. Other factors such as security were not in the scope of this project though

they were important to consider during development. For example, the only input from

the user is a map click (achieved by tapping the screen) for selecting a suburb or a tab

(for choosing to visualise data by suburbs or categories).

As the map is interactive, a user is able to select any suburb of interest to interrogate

is details. Figure 3.6 shows the details pertaining to Phaphamani suburb as an example

of a selected suburb. Figure 3.6 (a) shows the responses for the suburb from the entire

time span. As can be seen from the screenshot, only two categories were available from

that suburb, namely ‘smells alkaline’ and ‘brownish’, represented by red and brown colours

respectively as can be seen in the details part of the visualisation (the listview). Figure 3.6

3.4. ITERATION 2 43

Figure 3.5: (a) Screen displays features of the visualisation (b) Screen displaying overallresults for the past six days (c) Screen displaying results viewed by categories

(b) shows the details for visualising two days for the past two days. Only one category

is available within this window and consequently the suburb is coloured with a brown

colour. Also, since this is the only colour within the window, it constitutes 100% of the

data to be visualised. The next screenshot - Figure 3.6 (c) - shows results for a seven

day window. At this window, there are three responses split between the two available

categories. The category for water that is brownish is higher (at 66.6%) than that of the

category for water that smells offensive (at 33.3%). Both Figure 3.6 (b) and (c) have

the brown category being the highest category and so the suburb gets coloured with the

brown colour.

The zoom capability of the visualisation is depicted in Figure 3.7, where Figure 3.7 (a)

shows a default view of the map before zooming and Figure 3.7 (b) shows the map after

zooming. The zoom is achieved through the zoom controls (the +/- signs) or pinching in

and out on the map.

3.4. ITERATION 2 44

Figure 3.6: (a) Screen for displaying results for a suburb (b) Displaying results for thepast two days (c) Displaying results for the past six days

3.4.4 Analyse

To realise the third goal of this research, the developed prototype was evaluated by users.

The evaluation sought to establish if the prototype addresses some of the dimensions that

usability and UX for the visualisation prototype. citetsharp2011interaction describe UX

as

“... how people feel about a products and their pleasure and satisfaction when

using it, looking at it, holding it, and opening or closing it.”

It is also an important aspect to evaluate as it relates to users’ emotions and feelings as

they interact with a system (Sharp, Rogers, & Preece, 2011). Chittaro (2006) has shown

that evaluation is one of the key design considerations as it reveals usability problems

associated with the system. The evaluation process will be discussed in detail in Chapter

4. The entire processes involved in the evaluation - from setting up the evaluation through

the final process of presenting results from the evaluation - are discussed in the following

sections.

3.4. ITERATION 2 45

Figure 3.7: (a) Screen displaying a default suburb (b) Screen displaying a zoom view ofthe suburb

Experiment Design

Usability is a multidimensional property of a system which tests the ease with which user

interfaces are used (Nielsen, 2012). In evaluating usability, the investigator must decide

which dimension is most important for their context, users and goals. The dimensions

are:

� effectiveness - this ensures that the designed product performs as users desire, that

is, it meets the expectations of users

� efficiency - this ensures the product achieves the desired tasks in a minimum number

of steps

� safety/errors - users’ safety should be guaranteed and errors minimised or recover-

able

� utility - this aspect refers to the ability of a product to be as functional as users

would prefer, that is, provide expected functionality without restricting users to

very limited and undesirable options as they use the product

3.4. ITERATION 2 46

� learnability - this refers to the ease with which the product can be learnt by users

the first time they encounter it

� memorability - refers to the ability of the product to provide simple enough func-

tionality or functionality for users to be able to remember how to use after a while

without using it

Within this research, effectiveness was determined as most important because users should

be able to visualise data and draw distinctions between geographical and categorical

visualisations. Learnability was also deemed important since mobile users typically do

not get training on how to use the applications. So it was important to test that users

would be able to learn the visualisation on their own.

The DECIDE Framework

DECIDE is a framework which guides the process of designing and conducting evalua-

tions (Sharp et al., 2011). The framework caters for the entire process, from setting the

right goals for evaluation, setting the right questions to ask during the testing, through

presenting data (Sharp et al., 2011) as explained in the items below.

� Determine the goals of the planned evaluation. The goals for this evaluation were

to test usability of the system, particularly effectiveness, efficiency and learnability.

� Explore the broad research questions in order to meet users’ goals. Visualising

geographical and categorical data is the primary goals of this visualisation, as such,

questions that ensure users are able to achieve this goal must be explored.

� Choose the method for evaluation such that the questions asked in the above item

are adequately answered. Usability test seemed most appropriate for the evaluation

as users were to carry out some tasks on the visualisation UI (Sharp et al., 2011).

As evaluation gathers data from users to be analysed by the evaluator, appropriate

data gathering approaches need to be considered. Various data gathering techniques

were used, namely, observation, data recording and questionnaire. The combined

advantages of the techniques include:

– accurate capturing of questions and comments by both the users and the eval-

uators during the test

3.4. ITERATION 2 47

– playback of the recording later for data analysis

– identifying firsthand what users do as they interact with the system (in ob-

servation) as other data gathering methods may not be able to capture users’

actions and understanding users’ context (Sharp et al., 2011).

– responding to specific questions (in a questionnaire) defined for specific usabil-

ity test e.g. tasks testing for an interface.

� Identify the practical issues. Considering this before carrying out the evaluation

helps avoid hiccups that could jeopardise data gathering process and ultimately re-

sults from the evaluation. Initially a questionnaire and observation were the only

chosen techniques, but after considering the entire process of evaluation from gath-

ering data through to data analysis, it become clearer that audio recording would

come handy if users had questions to asks during the interviews. The questionnaire

was initially printed on paper so that users would be able to conduct the evalua-

tion from any place of their choice. After considering the inflexibility of hard-copy

questionnaires on the side of users, an online questionnaire seemed a better choice.

At this point, issues such as facilities and equipment to be used for the evaluation,

participants and schedules are considered. Another issue addressed was whether a

demonstration would be done for the participants. The researcher decided to leave

users to determine how to figure out the interface themselves so that learnability

would be assessed.

� Decide how to deal with the ethical issues. Standard procedures have been under-

taken to protect users. Rhodes University has a standard procedure for carrying out

evaluation that involves users. An approval was sought from the Hamilton Build-

ing Human Research Ethics Committee. The approval contains all the information

related to the evaluation. This information includes the description of the research,

the purpose of carrying out the evaluation, the type of participants that the evalua-

tion targets, information that will be provided to the participants and the method of

carrying out the evaluation. For instance, users need to understand the purpose of

the evaluation and have to give their consent before participating, and can opt out

of the evaluation at any point during the evaluation for any reason. For example,

the information given to participants and a consent form are shown in Listing A

while the rest of the documents related to ethics are included in the CD-ROM listed

as Appendix C.

� Evaluate, analyse, interpret and present the data. This item is crucial in that it

impacts all the decisions that will be made out of the evaluation exercise. It eval-

3.4. ITERATION 2 48

uates the evaluation itself, determining whether the methods used would produce

intended, valid and reliable results. Data analysis methods are also determined at

this stage. Due to the nature of questions developed from the second item. Both

quantitative and qualitative data analysis methods seemed plausible since usabil-

ity testing results would easily be interpreted quantitatively whereas user opinion

(usability and UX opinion) would best be analysed qualitatively.

User test

Nielsen (2014) describes how to develop task scenarios that should achieve the goals of

an evaluation. The scenarios tasks are said to engage users. These tips are by Nielsen

(2014, pp. 2):

� “make the task realistic” - avoid superfluous tasks would incite disbelief in the

user as the task could be something they would normally not do under normal

circumstances. This could be achieved by designing tasks that are not very specific

but instead general yet still verifiable by the researcher that the user was able or

unable to perform that task.

� “make the task actionable” - engage a user in an action to achieve a goal instead of

asking them to explain how they would achieve.

� “avoid clues and describing steps” - let a user determine how they would achieve a

goal without giving hints or tips. This is especially true in the mobile platform where

users do not usually receive trainings or detailed manuals as in the PC platform.

Users usually have to ‘find their way’ through the application and ultimately figure

out it functions.

Combining the DECIDE framework with some tips from (Nielsen, 2014) enabled the

design of the user evaluation to be as realistic as possible from the perspective of the user.

Some UX design principles were also adopted from Sharp et al. (2011). The user test

involved a scenario detailing a specific scenario. The action-oriented tasks then followed

the scenario. The tasks were set out as a way to engage a user with the application.

The evaluation tests how the users are able to achieve specific tasks/goals within the

application. The tasks were designed such that it would be clear to determine whether or

not the user was able to achieve a goal.

3.5. SUMMARY 49

The evaluation method - Controlled setting involving users - used for this research was

adopted from Sharp et al. (2011). It gave more control to the evaluator on what to let

users do though usually at the expense of users’ freedom to act naturally. It is often used

for usability testing.

Geographical and categorical data are the main data types used in this research work. For

this research, usability would be demonstrated if users could tell apart the geographical

and categorical components of the visualisation and also be able to see how they relate to

each other. The geographical component of the visualisation is explicitly represented by

the suburbs overlaid on Google Maps. The categorical component is represented by the

various responses from the water quality.

3.5 Summary

This chapter discussed the methodology followed throughout this research, specifically

the spiral model. The DECIDE evaluation framework was used to guide the evaluation

process. The framework guided the setting up of the user evaluation experiment. The vi-

sualisation prototype was designed based on findings from related work. Design principles

discussed in the related work were used to influence the design of the visualisation proto-

type. The implementation followed the design, discussing various tools and resources used

in producing the final prototype. Extensive use of APIs from the tools such as Google

Maps and Microsoft Azure have also proved to be useful - both for speedy development

and efficient use of the limited resources since they are specifically developed for mobile

applications. The next chapter presents and discusses the results of the user evaluation

for the prototype. The evaluation was designed to test for some aspects of usability (e.g.

effectiveness and learnability) and solicit participants’ experience with the visualisation.

Chapter 4

User Evaluation

4.1 Introduction

This chapter presents and discusses results obtained from the user evaluation. Evaluation

of the developed system was one of the goals of this research project. This involved testing

the visualisation prototype in terms of usability and UX. Evaluation determines how much

of the user needs the application is able meet. It is equally important to determine if the

needs of the users were met after development. The following sections describe results

obtained from the user evaluation as well as a discussion of the results of the prototype

designed and developed in Chapter 3.

4.2 Results

The approach employed for the evaluation was a qualitative approach as opposed to

quantitative approach. This approach allows for more detailed data to be gathered and

thus each user gets spends more time with with the application. This is advantageous

because users are expected to give feedback on their experience with the visualisation.

According to Nielsen (2000), 85% of usability problems are revealed by five (5) users

used in the user study, while the rest of the problems typically require large amounts

of resources so that large numbers of users are involved in the user study. A total of

eleven participants took part in the user evaluation. Although qualitative approach results

cannot be generalised as typically few participants are used in the evaluation, the benefit of

50

4.2. RESULTS 51

Table 4.1: Tasks(i) Find the categories with the highest and lowest responses.(ii) Find any suburb of interest and give the overall water quality sit-

uation in it?(iii) Give the water quality situation from the above suburb for the past

three (3) days.(iv) Find the suburb that has the highest incidence of water that smells

offensive.(v) What is the total number of reports from the above suburb?(vi) Which suburb reported the least responses of cloudy water in the

past ten (10) days?(vii) How many responses were reported in the above suburb?

the rich data gathered from participants outweighed the disadvantage of results not being

able to be generalised. Nine participants were from an ICT and an IS background while

one participant was from mathematical statistics background and the last participant was

from the biological sciences background. Owing to the participants’ expertise in the ICT

field, a significant number of suggestions were also provided. However, these have not

been incorporated into the visualisation due to time constraints. They will constitute

part of the material for the section on Future Work in the concluding chapter of this

thesis.

4.2.1 Usability

From the goals of usability discussed in Section 3.4.4, the tasks (from Question 1 of the user

study questionnaire) were designed to test for four of the six goals, these are effectiveness

and learnability. Memorability could not be tested for during the evaluation as it can only

be tested from subsequent evaluations from its nature. Similarly, safety could not be tested

for as the visualisation does not provide functionalities that would result in undesirable

consequences, e.g. accidentally deleting data and therefore requiring to ‘undo’ the action.

Table 4.1 shows a set of sub-questions for Question 1 of the questionnaire meant to be

tasks to be performed by participants. Each task was allocated marks depending on how

many parts were expected from a participant to be given as an answer to the task. During

scoring, one mark was allocated for each expected answer.

For Task (i), four marks were allocated as participants were expected to provide four

parts as an answer. The four expected parts were the highest and the lowest categories,

together with the respective number or percentage of responses corresponding to them.

4.2. RESULTS 52

Six participants (Participants 1-3 and 5-7) were able to give the required answers while

the rest (Participants 4 and 8-11) could not. Participants 4, 8, 9 and 10 used the wrong

screen to and so obtained wrong results, whereas Participant 11 used the correct screen

(the landing screen) but obtained results from the wrong part of the screen which was the

bar chart. This yielded a 55% success rate for the task. The question was testing if users

could make sense of the visualisation even before interacting with or exploring it beyond

the landing screen. The answers to the task were available from the landing screen before

the users were required to explore the rest of the interface. If participants could give the

correct answers, the information provided on the landing screen would prove effective,

hence proving the effectiveness goal of usability.

Task (ii) was allocated five marks as follow: one mark for clearly giving a suburb name and

four marks for giving the corresponding categories from the specified suburb. There was

no specific answer for this question as in task. The onus was on the researcher to identify

the suburb a participant chose and then check whether or not the given answer is correct.

For example, if a participant had chosen Fort England suburb, the water quality from that

suburb would be as follows: cloudy (9673/41.1%), brownish (8598/36.5%), smells alkaline

(4056/17.2%) and smells offensive (1206/5.1%). All participants were able to identify a

suburb, thereby realising that the map was interactive. However, only eight participants

associated the suburb details (displayed in the listview) with the chosen suburb. The

other three participants (Participants 2, 4 and 8) failed to relate the suburb they chose

to the categories and consequently got only one mark each. The question was testing

for interactivity as identifying a suburb required a user to tap on the map. It was also

testing for responsiveness as the details shown in the list vary according to suburbs while

the name of the selected suburb displays on the title bar.

Task (iii) was an extension of Task (ii). Participants were expected to slide the slider to a

three day window in order to give the correct answer. The same participants who correctly

answered Task (ii) were able to correctly answer this question. Additionally, Participant

4 was able to give the correct answer, resulting in nine (82%) participants successfully

achieving this goal. Only Participants 2 and 8 could did not respond correctly. Up to this

point in the evaluation, the questions were probing for information related to suburbs

(from the suburbs tab in the application). The expectation was that users would be

able to identify the various suburbs by tapping on the map and then view their details

(categories) from the listview located at the bottom of the screen.

The next set of tasks - Tasks (iv) to (vii) - were testing categories option of the application,

having suburbs as the details. One mark was allocated for each task. These tasks had

4.2. RESULTS 53

specific answers. The answer to Task (iv) was Vukani. All participants were awarded

a mark for this question even though some did not give Vukani as the answer. Various

reasons for awarding marks to ‘wrong’ answers are described in the next paragraph. The

next task - Task (v) - was an addition to Task (iv). It was designed to verify that

participants were able to read the details of the categories. The correct answer was

7242. Similarly, all participants were awarded a mark even though some participants

misunderstood the question. For Tasks (vi) and (vii), the correct answer were Extension

4 & Extension 5 and 2 respectively and all participants got the answers right.

During the scoring of the results, some participants did not give the expected answers

due to various reasons which are explained below. However, marks were awarded as the

answers they provided demonstrated that they could use the visualisation application

and make associations or establish relationships between the suburbs and their details

(categories) in the listview. The various reasons are as follows:

� interpreting the word ‘overall’ from Task (ii) to mean the results shown in bar chart

as ’Overall results’ - the researcher realised the wording of the question could have

caused the confusion and participants failed to understand the question rather than

the use of the visualisation.

� Interpreting the word ‘total’ from Task (v) as summing all responses for a particular

suburb/category - two participants (Participants 1 and 10) used calculators to add

all the categories in the listview and gave the total as the answer.

� Giving results of one tab with slider still in last tab position - the slider does not on

tapping another widget. The slider thumb was not reset deliberately to allow com-

parison between suburbs or categories within one tab, but could have been designed

to reset when a user moved from one tab to another. For example, Participants 2

and 4 did not reset the slider when they moved to the categories tab and got both

got Hlalani suburb as the answer. Participant 2 expected the slider reverts to ‘0

days selected’ upon tapping a different tab/suburb.

Results across all tasks yielded an overall average score of 79% from the scoring. Table

4.2 summarises the results. As can be seen from Table 4.2, five participants (Participants

1, 3 and 5-7) were able to perform all the specified tasks satisfactorily, that is, were able

to achieve 100% of the goals set out for them. A further three (Participants 9, 10 and

11) were able to perform well with a score of 78%. Participants 2 and 4 were in the mid

ranges (each at 56%) whereas Participant 8 struggled significantly to make effective use

4.2. RESULTS 54

Table 4.2: Scores for Results of the TasksParticipant (i) (ii) (iii) (iv) (v) (vi) (vii) Total Score(%)1 4 5 5 1 1 1 1 18 1002 4 1 1 1 1 1 1 10 563 4 5 5 1 1 1 1 18 1004 0 1 5 1 1 1 1 10 565 4 5 5 1 1 1 1 18 1006 4 5 5 1 1 1 1 18 1007 4 5 5 1 1 1 1 18 1008 0 1 0 1 1 1 1 5 289 0 5 5 1 1 1 1 14 7810 0 5 5 1 1 1 1 14 7811 0 5 5 1 1 1 1 14 78

Table 4.3: Question 2 - Usability Feedback(i) What stands out to you on the application? Explain your answer.(ii) Was the 10 days window enough? Would you want to look at results

from different days? How far back do you think you would want tosee?

(iii) Is the application easy or difficult to navigate? Can you explainyour answer?

(iv) Did you understand all the functions that the application provides?Did the labels and buttons do what you thought they would?

(v) Are colours used for representing categories well chosen? If not,can you suggest different colours?

(vi) Is the information and details displayed on the application sufficientor confusing (e.g. labels, details from data, etc.)?

(vii) Is there any other information or details you would like to see avail-able on the visualisation?

of the application, getting a score of just (28%). Reasons for these scores are discussed in

the Section 4.3. Section 4.3 also gives a discussion of the reasons reported by participants

and those observed by the researcher during the evaluation sessions.

4.2.2 Usability Feedback

This part of the evaluation investigated how usable the visualisation was from the par-

ticipants’ perspective. The questions were drafted to better understand how and why

they were able to cope (or not cope) with the tasks asked. The responses from this part

required users to give their opinion on the overall system and on specific aspects of the

system such as the time slider and colours used in the visualisation. Question 2 (i) was

4.2. RESULTS 55

Table 4.4: Responses for Question 2 (i) - Usability FeedbackParticipant Response1 ease to use, find info by tap2 ease to understand; no landscape mode, map in categories tab un-

necessary, highlight suburb on clicking on the list3 likes map, ease to use, interactive; no suburb names4 ease of selecting suburb, colouring in suburbs5 interactive, details shown, colour helps better understanding, labels

clear, easy to navigate6 simple layout, intuitive category selection7 Google Maps good choice, UI with Google maps8 suburbs, touch/tap capability, Google maps integration9 every suburb is affected by the water problem10 map with suburbs11 Google Maps, ease to navigate/select area of interest

investigating what participants found obvious or stood out within the application, espe-

cially in terms of usability. The responses from participants are summarised in Table

4.4. Responses varied from the application being easy to use, easy to understand, easy

to select suburbs, easy to navigate, interactive to having simple layout, integration with

Google Maps, colours helping to understand. Participant 9’s response was “every suburb

is affected by the water problem” and this response might have been a failure to contex-

tualise the question. Participants 1, 2, 3, 4 and 5 noted that the visualisation was easy

to use or navigate. Participants 3, 7, 8, 10 and 11 liked the inclusion of the maps.

Question 2 (ii) was meant to assess whether or not the default ten days window provided

by the slider was enough. Eight participants gave explicit responses, equally split between

being ‘enough’ and ‘not being enough’. The remaining three (Participants 4, 6 and 7)

preferred to have users choose their own time span by having to pick from two dates. A

summary of the responses is given in Table 4.5.

Question 2 (iii) investigated the ease or difficulty of use. Six participants (Participants 1,

2, 3, 4, 5 and 8 ) stated it was easy to use, easy to navigate or intuitive. Six (Participants

3, 5, 7, 9, 10 and 11) reported it was difficult at first, but it became easy after a while

or became easier to use after some time. Participant 2 reported the application might

present challenges with users with little exposure to technology. Participant 8 stated that

the application was difficult to use. The participant also gave suggestions on how to

improve the UI. For example, suggestions such as customisable time picker, provision of

a search bar and a drop down menu on the map were made as shown in Table 4.6. The

4.2. RESULTS 56

Table 4.5: Responses for Question 2 (ii) - Usability FeedbackParticipant Response1 not enough; would prefer to go up 30 days back2 enough for short term; but also provide for X years3 very short; provide in log scale (e.g. 1 day, 3 days, 1 week, 2 weeks,

month as user increases time)4 let it be customisable - let user choose two between dates5 enough without redundancy6 let it be customisable - let user choose two between dates7 customisable - let user choose two between dates8 enough9 not enough; up to 30 days back10 not enough up to 30 days back11 enough; but get current data - not previous one

table summarises1 the responses from participants.

Nine participants (Participants 1, 2, 4-6, 8-11) responded that they were able to under-

stand the functions provided by the visualisation in response to Question 2 (iv). The

other remaining participants (Participant 3 and 7) identified problems with certain in-

terface aspects. For example, Participant 3 did not think the suburb name updated in

the title bar was logical. Participant 3 also commented that instruction on how to use

the slider was not clear or was misleading. Participant 7 similarly had confusion with the

time slider and sought clarification during the test. A summary of responses is given in

Table 4.7.

The use of colours, in Question 2 (v), was reported to be well chosen by eight participants

(besides Participants 2 and 3). Of the remaining three, Participant 2 had no opinion.

Participant 3 suggested grey instead of blue colour be used for the cloudy category (but

also pointed out they thought it did not matter). Participant 3 also warned about the use

of the colour red for the category ‘smells alkaline’, indicating that the colour makes this

category seem more serious than the others and suggested the use of a different colour

“to maintain an equal footing between the categories”. Participant 6 responded that

the colours made it easier to use the visualisation, but commented that the extent of the

problem was not clearly indicated by colours. Table 4.8 gives a summary of the responses.

For Question 2 (vi) Participants 1, 2, 4, 7, 8 and 9 responded that the information and

details displayed on the visualisation were sufficient. Two participants (Participants 2 and

5) indicated that the instruction/tip on the slider was not clear. Participant 1 reported

1All responses from participants are available in the CD-ROM accompanying this thesis.

57

Table 4.6: Responses for Question 2 (iii) - Usability FeedbackParticipant Response1 very easy, all info there; just tap icon to get2 may present difficulties with novice users3 easy to navigate, logical4 very intuitive, easy to navigate; slider difficult to use5 easy to navigate; need learn to use it first6 very intuitive due to separation of suburbs and categories7 bit difficult to use at first, but very interesting after some time8 difficult; provide location search/drop down, allow custom date

pickers9 difficult to use at first, but easy to navigate after some time10 confusing at first, but easier after some time11 easy to use due to tabs; slider difficult to use, name updated in title

bar not easy to see, need familiarise yourself then it becomes easierto use

Table 4.7: Responses for Question 2 (iv) - Usability FeedbackParticipant Response1 yes2 yes but labels could be clearer3 logical; tapping a suburb from a listview could select it on a map,

suburb name on title bar not logical, ‘0 days selected’ misleading4 yes5 yes6 very simple layout - minimal widgets7 did not understand categories tab, selected days8 yes9 yes10 yes11 yes

4.2. RESULTS 58

Table 4.8: Responses for Question 2 (v) - Usability FeedbackParticipant Response1 yes - they are universally common2 no opinion3 use instead colour instead of blue for cloudy, red gives more em-

phasis so different color4 great colours5 excellent colour choice6 opacity not distinct7 well chosen8 well chosen9 well chosen10 well chosen11 well chosen

Table 4.9: Responses for Question 2 (vi) - Usability FeedbackParticipant Response1 not confusing2 slider instruction not clear3 sufficient4 sufficient5 slider instruction not clear6 simple data, easy to understand7 sufficient8 sufficient9 sufficient10 labels initially confusing but got clearer as time passed11 percentage of responses was initially confusing

nothing was confusing, while Participant 10 reported that it was initially confusing but got

clearer after spending more time interacting with the application. Participant 11 reported

that the ‘percentage of responses’ was initially confusing. A summary of responses for

Question 2 (vi) is given in Table 4.9.

The last question - Question 2 (vii) - explicitly sought additional features that users may

want to see in the visualisation. As Table 4.10 shows, the following suggestions were made

by participants. Four participants (Participant 1, 5, 9 and 10) did not want any additional

features within the visualisation. Two participants (Participants 3 and 11) preferred to

have suburbs coloured before interacting with the map. Participant 2 wanted to know

about the number of residences who live a given suburb, whereas Participant 6 wanted

to know about the severity of the problem. Participant 7 wanted data for other services

4.2. RESULTS 59

Table 4.10: Responses for Question 2 (vii) - Usability FeedbackParticipant Response1 no additional details2 number of residences3 overall colouring even before selecting suburbs/categories4 could expand listview on tapping because some names are too long5 sufficient6 severity of the problem visually7 have other data (e.g. electricity)8 laymen terms used to represent data9 no additional details10 no additional details11 overall colouring even before selecting suburbs/categories

(e.g. electricity, available from the second activity of the application) to be available for

visualising. Participant 4 wanted expandable list items, citing that they would like to see

the names of the suburbs since some were too long to be displayed in the space available

for display in the listview. Participant 8 commented “layman terms used to represent

data” but this did not make sense to the researcher as no examples were given for which

terms they wanted. Some suggestions from this question will constitute future work that

may arise from this research work.

4.2.3 User Experience

Question 3 - Interacting with the visualisation - formed the last part of the evaluation. As

with the usability feedback, there were no correct or incorrect answers when asking about

UX. The questions were investigating feelings or emotions invoked by the visualisation.

The results indicate a number of features that participants liked about the visualisation.

These were the use of Google Maps (7)2 within the visualisation, interactivity (4), selection

of suburbs (3), slider (1), tabs (1) and drop down menu for categories (1).

Regarding unpleasant features, four participants (Participants 1, 7 , 9 and 10) said there

were none; three (Participants 4, 6 and 11) commented on the slider, saying that it was

difficult to use, small and unintuitive. Participant 4 was not pleased with the clickable list

items that unfortunately did not reveal any more details e.g. a suburb names (especially

for those names that were too long to display fully in the list). Participants also provided

suggestions on how to improve the unpleasant or missing features within the visualisation

2numbers in the brackets indicate the number of participants who listed the given feature as pleasant

4.2. RESULTS 60

Table 4.11: Question 3 - User Experience(i) Were there any aspects of the application that you found pleasant

(e.g. satisfying, helpful, provocative, etc.)?(ii) Were there any aspects of the application that you found unpleasant

(e.g. frustrating, unpleasant, boring, etc.)?(iii) Is there anything you would like to comment about the MobiSAM

system?

application, for example, to provide a ‘Help or Tips’ section or icon to quickly guide users

on the use of features.Participant 8 commented “The fact that i[sic] move the data on full

screen”. This response could not be understood by the researcher.

The last question of the evaluation asked about overall feedback on the MobiSAM system.

Responses, as can be seen from Table 4.12, vary from being informative, easy to use, nice

to user friendly. However, Participants 2, 8 and 9 provided no answer to this question.

Suggestions given by Participants 3, 6, 7 and 10 were highlighted and discussed in the

usability responses earlier in the previous section. The following section discusses the

results of the evaluation.

Table 4.12: Results from Question 3 - User Experience

Participant Pleasant Unpleasant Comments

1 all info there no interesting, easy to

use, very informative,

fascinating how info

revealed

2 map re-navigating from

one tab to another to

analyse a category

no

3 selecting suburb,

slider

make suburb names in

the categories expand-

able

consider other colours

for instruction as it

make it look like an er-

ror

4 selecting suburb, map

interactivity satisfying

slider finicky, list

items clickable but do

not do anything

very nice and user

friendly app

Continued on next page

4.3. DISCUSSION 61

Table 4.12 – continued from previous page

Participant Pleasant Unpleasant Comments

5 map, interactivity,

colours, professional

looking app, flows

well

not help button/icon,

provide list of suburbs

to select from, im-

prove responsiveness

great app, ready for

deployment

6 Google Maps use

leverages user famil-

iarity so becomes easy

to use

slider small, unintu-

itive, difficult to use

could overlay suburb

name on zoom for easy

of search

7 Google Map, suburb

details/problems

no implement other ser-

vices, include weekly

and monthly timelines

8 option bar, map move data on full

screen

no

9 easy to visualise no no

10 yes, tab labels give a

hint

no include help icon/sec-

tion for user friendli-

ness

11 touch interface is

pleasant, no typing -

just tap

slider difficult to use very relevant, user

friendly UI

4.3 Discussion

User evaluation results indicated an overall success for the visualisation prototype with

regard to usability and UX. This indicates that the prototype is usable (to a satisfactory

degree - 79% from the usability testing) and users find it an easy to use, nice and user

friendly application to use. As shown in Tables 4.2, 4.4 - 4.10 and 4.12, the overall

evaluation yielded positive feedback. Majority (79%) of the usability tasks were achieved

by participants. Participants also gave an overall positive feedback on the usability of the

visualisation and UX. Most (nine out of eleven) of the participants were from the ICT

field and as such, gave a lot of feedback influenced by their expertise in the field. Most of

the problems they found were interface related as usability tests are meant to test for such

problems. The usability success was attributed to the prototype’s ease of use, its simple

4.3. DISCUSSION 62

UI and the integration of Google Maps amongst other factors. The following sections

discuss the results for usability and UX respectively.

4.3.1 Usability Feedback

As described in Section 3.4.4, the user test was designed to test the usability of the

visualisation. The average result for usability (from Question 1 of the questionnaire)

was 79%. This score indicates that most of the set tasks were successfully performed

by participants. The results given in Table 4.2 indicate a pattern of increasing success

in tasks as participants spent more time with the visualisation. This pattern proves

that the prototype was learnable. The participant (Participant 8) who struggled with

the tasks was, interestingly enough, from the ICT background. From the observations

made by the researcher, the participant was least interested in the evaluation and the

participant’s performance was not due to lack of familiarity with technology. It is also

worth noting that during the evaluation, no demonstration was made by the researcher to

participants. Participants were instead asked to launch the application from the mobile

phone and familiarise themselves with it, then start the user evaluation when they felt they

were ready. Five participants reported that the integration of Google Maps within the

visualisation improved intuitiveness since many users are already familiar with Google

Maps. Interactivity with the map was also highlighted as a beneficial and convenient

aspect that improved UX. Another aspect that appealed to users was the simplicity of

the visualisation.

Just over half of the participants successfully achieved the fist task - Task (i) - while

the rest of the tasks were performed satisfactorily, that is, all participants almost got

all questions correct. One reason that stood out from the researcher’s observation was a

short time users spent familiarising themselves with the application. This was evidenced

by comments made later during the test when participants reported they it was easier

to use the application after spending more time on it. During the evaluation, one other

shortcoming from the visualisation was the lack of feedback to users when they were

waiting for something to load. As results were being fetched from the online database, no

indication of the progress was made to users and that they need to wait a few seconds.

Most users would tap a screen and within a short period, as results were being fetched

from the server, tap on a different place. This resulted in slight frustration, especially

early on during the evaluation before they realised that they had to wait a few seconds.

The frustration was a limitation of the application to provide some form feedback to the

4.3. DISCUSSION 63

user (through implementing a progress bar or icon) as results are being fetched from the

server.

Three participants (4, 6 and 11) complained about the small size of the time slider.

Participant 4 described the slider as being “a bit finicky”. Participants were not able to

easily move it to the exact number of days they desired. As such, they suggested that a

better alternative would be allowing a user to pick dates in some way. The time element

(abstracted by the slider) within the application was also slightly confusing. Participants

1, 3 and 4 sought clarification as to what it meant or confirmation as to whether it was a

time period for the past N days where N is the number of days selected from sliding the

time slider.

Nine participants also appreciated the interactivity of the Google Maps and the overlay

of suburb boundaries, citing the ease of selecting suburbs of interest. However, three

participants (Participants 2, 8 and 11) suggested that since the suburbs were not labelled

on the map, it would be ‘nice’ to have a list (possibly a drop-down menu) where a user

would quickly just select a suburb of interest and have it shown/coloured on the map.

Participant 5 highlighted the importance of having a help or tip functionality so that

users would use it as a guide on how to use the visualisation.

One other feature that caused confusion was the two tabs in the visualisation. As described

in Chapter 3, the Suburbs tab was designed to be used to visualise the various suburbs

while the categories tab was meant for viewing the various categories. On each tab, say

the suburbs tab, a user would be expected to select (by tapping) a suburb to view its

details. The details for the suburb are the various categories of a suburb. While on a

categories tab, a user would be expected to view the category by selecting the appropriate

one, and then view how the various suburbs are affected by the chosen category. This

relationship was not obvious. For example, Participant 2 was not able to distinguish

between getting information pertaining to a particular suburb and that pertaining to

a particular category. As can be seen from the screenshot in Figure 3.5, a red colour

was used to highlight the instruction of interacting with the application. Participant

3 acknowledged this, but indicated that it looks like an error message and suggested a

different colour be used. Participant 11 was not aware that on launching the visualisation,

the results displayed were for all suburbs and for the entire timespan of the dataset and

that the bar chart at the top summarised the entire dataset.

4.4. SUMMARY 64

4.3.2 User Experience

Some results for UX indicated that participants expressed positive feelings and emotions

towards the visualisation. For example, Participant 1 reported that the application was

interesting and fascinating in terms how much information it revealed while Participant

4 described the visualisation as nice and user friendly. Most of the pleasant experiences

were attributed to the features users liked. For example, Google Maps was liked by seven

participants (Participants 1, 3, 5, 7, 8, 10 and 11) while others participants (e.g. Partic-

ipants 4, 5 and 11) liked the interactivity of the application and the map. Participants

reported a level of satisfaction as they interacted with the application and indicated some

were able to acknowledge the ease of its navigation and simplicity. There was however one

participant (Participant 2) was not fond of the navigation, citing it was a bit difficult to

manoeuvre through the application. The participant commented that they only realised

the map was interactive towards the end of the evaluation.

4.4 Summary

This chapter gave a detailed discussion of the evaluation results of the visualisation.

The results from the usability testing highlight that it is relatively simple to utilise the

visualisation application. The simple UI and the integration of Google Maps proved

to be desirable features for the visualisation. However, a few interface problems were

highlighted, e.g. the difficulty using the time slider. Comments and suggestions on

how the visualisation could be improved were also obtained from the participants. As

already indicated, these would be part of Section 5.3 - Future Work. The results for the

evaluation thus provide evidence that the prototype is effective for visualising datasets

for large geographical and categorical data. The next chapter concludes the thesis. It

summarises the entire thesis, highlighting important elements of the research process and

findings.

Chapter 5

Conclusion

5.1 Overview

The aim of this research was to investigate visualisation and interaction techniques for

large datasets for the mobile platform. The specific goals were also specified in Chapter 1

and how they were achieved was detailed in the chapters on related work, methodology and

user evaluation. It has been found that visualisation and interaction techniques largely

depend on the type of data used and the type of target device for the visualisation.

That said, the visualisation prototype indicates that visualisations for large datasets can,

indeed, be created for mobile phones. It has also showed that geographic and categorical

datasets can be visualised on mobile phones. Making use of APIs (e.g. Google Maps

API) not only saved development time but improved performance as well, as APIs are

optimised for performance and therefore take cognizance of the limited resources of mobile

devices. It also improved UX as users are already familiar with Google Maps.

5.2 Research Goals Revisited

The research goals stated in the introductory chapter of this thesis were all achieved. A

summary of each goal is given below:

� to review mobile visualisations and interactions techniques for large datasets.

Findings from related work indicated that limited resources in the mobile platform

are a major hindrance to visualisations in the mobile platform;

65

5.3. FUTURE WORK 66

� to develop visualisations based on the findings from related work.

Sections 3.3 and 3.4 discussed the design and development of the visualisation pro-

totype, iteratively applying the design guidelines from the the findings on related

work. Extensive use of APIs specifically developed for the platform improved the

performance and functionality of the developed visualisation prototype; and

� to evaluate the visualisations for effectiveness and usefulness amongst other factors.

Section 4.2 presented the results of the evaluation. The results of the evaluation were

discussed in Section 4.3. They results were positive - indicating that visualisations

for large datasets can be provided on mobile phones. Summing the results for

this portion of the thesis, it can be concluded that visualisation and interaction

techniques for large datasets on the mobile platform can be effective and useful for

providing information to users in a simple and clear manner.

5.3 Future Work

Findings from this research can be used to inform later research in a number of ways.

Implementation was done on the Android platform. This only narrows the use of APIs

and tools that enhanced development of the prototype only to the Android platform. It

would be worthwhile to extend the work to other platform such as iOS and Windows

and compare usability and UX across platforms. Time constraints could also not allow

incorporation of suggestions made by the participants during the evaluation of the visual-

isation. These include modification of features such as the time slider, more information

revealed on clicking items and colouring of the suburbs upon landing on the tab screens.

Future work could also explore how users may make their own determination on what to

visualise. Other data types, for instance text-based data, could also be interesting to use

in developing visualisations and on other existing platforms such as iOS and Windows.

References

ACM. (2014). The 2012 ACM Computing Classification System. Retrieved from http://

www.acm.org/about/class/2012 (Accessed on 10 October 2014)

Akbari, M., Alfut, A., Balaz, A., Bloor, R., Bradley, D., Buttner, M., et al. (2014). Mobile

developer’s guide to the galazy. Enough Software.

Android Developers. (2014). ADT Plugin. Retrieved from http://developer.android

.com/tools/sdk/eclipse-adt.html (Accessed on 11 Sep 2014)

Apple. (2013). iPhone 5s. Retrieved from https://www.apple.com/za/iphone-5s/

specs/ (Accessed on 23 May 2014)

Ballagas, R., Borchers, J., Rohs, M., & Sheridan, J. G. (2006). The smart phone: a

ubiquitous input device. Pervasive Computing , 5 (1), 70–77.

Baudisch, P., Good, N., & Stewart, P. (2001). Focus plus context screens: combining

display technology with visualization techniques. In Proceedings of the 15th annual

ACM symposium on User interface software and technology (pp. 31–40).

Baudisch, P., & Rosenholtz, R. (2003). Halo: a technique for visualizing off-screen objects.

In Proceedings of the SIGCHI conference on Human factors in computing systems

(pp. 481–488).

Boehm, B. W. (1988). A spiral model of software development and enhancement. Com-

puter , 21 (5), 61–72.

Callegaro, M. (2013). Do you know which device your respondent has used to take your

online survey? Survey Practice, 3 (6). Retrieved from http://www.surveypractice

.org/index.php/SurveyPractice/article/view/250/html

Card, S. K., Mackinlay, J. D., & Shneiderman, B. (1999). Readings in information

visualization: Using vision to think. Morgan Kaufmann.

Carr, D. A. (1999). Guidelines for designing information visualization applications. Pro-

ceedings of ECUE , 99 , 1–3.

Chittaro, L. (2006). Visualizing information on mobile devices. Computer , 39 (3), 40–45.

Dix, A., Finley, J., Abowd, G., & Beale, R. (2004). Human-computer interaction. Pearson

Prentice Hall.

67

REFERENCES 68

Fling, B. (2009). Mobile design and development: Practical concepts and techniques for

creating mobile sites and web apps. O’Reilly.

Fry, B. (2009). Visualizing data. O’Reilly.

Google Developers. (2014). Getting Started. Retrieved from https://developers

.google.com/maps/documentation/android/start (Accessed on 11 Sep 2014)

GSMA. (2013). The mobile economy sub-saharan africa 2013. Retrieved from

http://www.gsmamobileeconomyafrica.com/Sub-Saharan%20Africa ME Report

English 2013.pdf (Accessed on 20 October 2014)

Heer, J., Bostock, M., & Ogievetsky, V. (2010). A tour through the visualization zoo.

Communications of the ACM , 53 (6), 59–67.

Ishimaru, S., Kunze, K., Kise, K., Weppner, J., Dengel, A., Lukowicz, P., & Bulling, A.

(2014). In the blink of an eye: Combining head motion and eye blink frequency for

activity recognition with google glass. In Proceedings of the 5th Augmented Human

International Conference (pp. 15:1–15:4). ACM.

ITU. (2014). The world in 2014: ICT Facts and Figures. Retrieved from http://www.itu

.int/en/ITU-D/Statistics/Documents/facts/ICTFactsFigures2014-e.pdf

(Accessed on 26 May 2014)

Kutter, O., Shams, R., & Navab, N. (2009). Visualization and gpu-accelerated simu-

lation of medical ultrasound from ct images. Computer methods and programs in

biomedicine, 94 (3), 250–266.

Mirosoft Azure. (2014). Get started with Mobile Services. Retrieved

from https://azure.microsoft.com/en-us/documentation/articles/

mobile-services-android-get-started/ (Accessed on 11 Sep 2014)

Mouton, C., Sons, K., & Grimstead, I. (2011). Collaborative visualization: current

systems and future trends. In Proceedings of the 16th International Conference on

3D Web Technology (pp. 101–110).

News24. (2014). News24 Elections (Version 1.5.2) [Mobile Application Soft-

ware]. Retrieved from https://play.google.com/store/apps/details?id=za

.co.elections24 (Accessed on 22 May 2014)

Nielsen, J. (2000). Why You Only Need to Test with 5 Users. Retrieved from http://

www.nngroup.com/articles/why-you-only-need-to-test-with-5-users/ (Ac-

cessed on 24 Sep 2014)

Nielsen, J. (2012). Usability 101: Introduction to usability. Retrieved from http://www

.nngroup.com/articles/usability-101-introduction-to-usability/ (Ac-

cessed on 24 Sep 2014)

Nielsen, J. (2014). http://www.nngroup.com/articles/task-scenarios-usability-testing/.

Retrieved from http://www.nngroup.com/articles/task-scenarios-usability

REFERENCES 69

-testing/ (Accessed on 24 Sep 2014)

Paelke, V., Reimann, C., & Rosenbach, W. (2003). A visualization design repository for

mobile devices. In Proceedings of the 2nd International Conference on Computer

Graphics, Virtual Reality, Visualisation and Interaction in Africa (pp. 57–62).

Partridge, K., Chatterjee, S., Sazawal, V., Borriello, G., & Want, R. (2002). TiltType:

accelerometer-supported text entry for very small devices. In Proceedings of the 15th

annual ACM symposium on User interface software and technology (pp. 201–204).

Peng, W., Ward, M. O., & Rundensteiner, E. A. (2004). Clutter reduction in multi-

dimensional data visualization using dimension reordering. In Information visual-

ization, 2004. infovis 2004. ieee symposium on (pp. 89–96).

Pombinho, P., Carmo, M. B., & Afonso, A. P. (2009). Evaluation of overcluttering pre-

vention techniques for mobile devices. In Proceedings of the 2009 13th International

Conference on Information Visualisation (pp. 127–134).

Rhyne, T. M., & MacEachren, A. (2004). Visualizing geospatial data. In ACM SIG-

GRAPH 2004 Course Notes. ACM.

Robbins, D. C., Cutrell, E., Sarin, R., & Horvitz, E. (2004). ZoneZoom: Map navigation

for smartphones with recursive view segmentation. In Proceedings of the Working

Conference on Advanced Visual Interfaces (pp. 231–234).

Rosenholtz, R., Li, Y., Mansfield, J., & Jin, Z. (2005). Feature congestion: a measure

of display clutter. In Proceedings of the SIGCHI conference on Human factors in

computing systems (pp. 761–770).

Ross, G. (2008). Visualization: A magic table. Retrieved from http://magic

-table.googlecode.com/svn/trunk/magic-table/google visualisation/

example 1.html (Accessed on 29 May 2014)

Rukzio, E., Broll, G., Leichtenstern, K., & Schmidt, A. (2007). Mobile interaction with the

real world: An evaluation and comparison of physical mobile interaction techniques.

In Ambient intelligence (pp. 1–18). Springer.

Samsung. (2013). Samsung Galaxy S4. Retrieved from http://www.samsung

.com/africa en/consumer/mobile-phone/mobile-phone/smart-phone/

GT-I9500ZWEXFA-spec (Accessed on 23 May 2014)

Sayyed, A. (2012). System Development Life Cycle (SDLC). Retrieved from http://

ars-mis5.blogspot.com/2010/04/sad.html (Accessed on 30 October 2014)

Sharp, H., Rogers, Y., & Preece, J. (2011). Interaction design: beyond human-computer

interaction. Wiley.

Shneiderman, B. (1996). The eyes have it: A task by data type taxonomy for information

visualizations. In Proceedings of 1996 IEEE Symposium on Visual Languages (pp.

336–343).

REFERENCES 70

Shneiderman, B. (2004). Designing for fun: how can we design user interfaces to be more

fun? Interactions , 11 (5), 48–50.

Soros, G., Seichter, H., Rautek, P., & Groller, E. (2011). Augmented visualization with

natural feature tracking. In Proceedings of the 10th International Conference on

Mobile and Ubiquitous Multimedia (pp. 4–12). ACM.

Thinyane, H., & Coulson, D. (2012). Mobisam: Mobile social accountability monitoring

in south africa. Proceedings of M4D 2012 28-29 February 2012 New Delhi, India,

28 (29), 170.

Van Tonder, B., & Wesson, J. (2008). Using adaptive interfaces to improve mobile map-

based visualisation. In Proceedings of the 2008 annual research conference of the

south african institute of computer scientists and information technologists on it

research in developing countries: riding the wave of technology (pp. 257–266).

Wang, Y., Huynh, G., & Williamson, C. (2013). Integration of google maps/earth with

microscale meteorology models and data visualization. Computers & Geosciences ,

61 , 23–31.

Ware, C. (2013). Information visualization: Perception for design. Morgan Kaufmann.

Wigdor, D., & Balakrishnan, R. (2003). TiltText: using tilt for text input to mobile

phones. In Proceedings of the 15th annual ACM symposium on User interface soft-

ware and technology (pp. 81–90).

Yoo, H. Y., & Cheon, S. H. (2006). Visualization by information type on mobile device.

In Proceedings of the 2006 Asia-Pacific Symposium on Information Visualisation

(Vol. 60, pp. 143–146). Australian Computer Society, Inc.

Zellweger, P. T., Mackinlay, J. D., Good, L., Stefik, M., & Baudisch, P. (2003). City

lights: contextual views in minimal space. In CHI’03 extended abstracts on Human

factors in computing systems (pp. 838–839).

Appendix A

User Evaluation

A.1 Information Sheet

Information to participants

Title of Research: Mobile Visualisation Techniques for Large Datasets

Principal Researcher: Motebang Lebusa

Supervisors: Prof. Hannah Thinyane and Mrs. Ingrid Siebrger

Purpose of research

The purpose of this research project is threefold:

� to evaluate mobile visualisation and interaction techniques for large datasets;

� to develop visualisations based on findings from related work; and

� to evaluate the visualisations for usability and user experience.

The user study forms part of the last research goal evaluation. Its main purpose is to as-

sist in understanding usability (and discovering associated problems) and the experience

users have as they interact with the application.

71

A.2. CONSENT FORM SIGNED BY PARTICIPANTS 72

User study

The application that is to be evaluated in this study uses sample data from a real world

project called MobiSAM. MobiSAM is being implemented in the Makana Municipality

(here in Grahamstown). The purpose of that project is to support the community to

participate in local government by reporting problems related to service delivery, and

providing a way for the municipality to communicate planned / unplanned service delivery

outages to the community. Registered users report on their service delivery issues and

responses are summarised by suburbs. The municipal authorities access the data/reports

and then take appropriate action to resolve them.

The system currently requests feedback around the provision of water, sanitation, roads,

and electricity.

Within the application, the various responses are represented by different colours. These

colours are used to colour the suburbs, such that the category that has the highest number

of responses is used as the overall colour of the suburb.

Recordings

For the duration of the user study, audio recordings will be made as a way of accurately

capturing data.

A.2 Consent Form signed by Participants

CONSENT FORM




Project Aims

The aims of this research project are threefold:

1. to evaluate mobile visualisation and interaction techniques for large datasets;

2. to develop visualisations based on work from related work; and

3. to evaluate the visualisations for usability and user experience.

A.3. EVALUATION QUESTIONNAIRE 73

Purpose of User Study

The purpose of this user study is to evaluate a visualisation system for categorical and

location based information. In doing so it seeks to assist in understanding the usability

(and discovering associated usability problems) and the experience users have as they

interact with the application.

� I have received information about this research project.

� I understand the purpose of the research project and my involvement in it.

� I understand that I may withdraw from the research project at any stage.

� I understand that participation in this user study is done on a voluntary basis.

� To the best of my knowledge I have no physical impediments that will stop me from

completing this study.

� I understand that while information gained during the study may be published, I

will not be identified and my personal results will remain confidential.

Name of participant.....................................................................................................

Signed......................................................... Date............................................................

I have provided information about the research to the research participant and believe

that he/she understands what is involved.

Researchers signature and date.................................................................................

A.3 Evaluation Questionnaire





Scenario

Your town Grahamstown has constant water problems. You’ve been having

terrible water quality in your suburb for the past two weeks and would like

to know extent of the problem in the whole town. Launch MobiSAM app

and find out more information and then perform the tasks given below.

1. Tasks

i) Find categories with the highest and least responses.

ii) Find any suburb of interest and give the overall water quality situation in it?

Suburb name:

Overall water quality:

iii) Give the water quality situation from the above suburb for the past three (3) days.

iv) Find the suburb that has the highest incidence of water that smells offensive.

v) What is the total number of reports from the above suburb?

vi) Which suburb reported the least responses of cloudy water in the past ten (10) days?

vii) How many responses were reported in the above suburb?

2. Usability

i. What stands out to you on the application? Explain your answer.

ii. Was the 10 days window enough? Would you want to look at results from different

days? How far back do you think you would want to see?

iii. Is the application easy or difficult to navigate? Can you explain your answer?

iv. Did you understand all the functions that the application provides? Did the labels

and buttons do what you thought they would?

v. Are colours used for representing categories well chosen? If not, can you suggest

different colours?

vi. Is the information and details displayed on the application sufficient or confusing (e.g.

labels, details from data, etc.)?

vii. Is there any other information or details you would like to see available on the

visualisation?


3. Interacting with visualisation

i. Were there any aspects of the application that you found pleasant (e.g. satisfying,

helpful, provocative, etc.)?

ii. Were there any aspects of the application that you found unpleasant (e.g. frustrating,

unpleasant, boring, etc.)?

iii. Is there anything you would like to comment about the MobiSAM system?

End of Questionnaire

Appendix B

Useful Information

B.1 Code Listing for Point in Polygon method

Listing B.1: pnpoly method listing

/*

* The method determines whether or not a point specified by the given coordinates

* lies inside the boundary of given by the array of the x−y coordinates

*/

public boolean pnpoly ( int npol , float [ ] xp , float [ ] yp , float x , float y )

{int i , j ;

boolean c = false ;

f o r ( i = 0 , j = npol−1; i < npol ; j = i++) {i f ( ( ( yp [ i ] > y ) != ( y < yp [ j ] ) ) &&

( x < ( xp [ j ] − xp [ i ] ) *

( y − yp [ i ] ) / ( yp [ j ] − yp [ i ] ) + xp [ i ] ) )

c = ! c ;

}r e turn c ;

}

B.2 Listing for all suburbs used in the visualisation

Listing B.2: List of suburbs used in the visualisation

Cradock Heights

Eluxolweni

76

B.3. LISTING FOR CATEGORIES USED IN THE VISUALISATION 77

Extension 4 & Extension 5

Extension 6 & Lingelihle

Extension 7 ( formal & informal )

Extension 8 , Extension 9 & Transit Camp

Fingo Village

Fort England

Grahamstown Central

Hill 60

Hlalani

Hoegenoeg

Joza & Extension 1

Kingswood

Mnandi , Extension 2 & Extension 3

Newtown , Nadncame & Silverton

Oatlands

Oatlands North

Phaphamani

Somerset Heights

Sunnyside

Tantyi , Xolani & Zolani

Vukani

West Hill

B.3 Listing for categories used in the visualisation

Listing B.3: List of categories used in the visualisation

cloudy

brownish

smells offensive

smells alkaline

Appendix C

Accompanying CD-ROM

The accompanying CD-ROM contains the following:

MotebangLebusa.pdf This thesis in pdf format.

References Electronic copies of some of the references used for this thesis.

SourceCode The source code of the visualisation developed in this research project.

Ethical Clearance Documents Electronic copies of documents submitted to the De-

partmental Human Research Ethics Committee for user evaluation.

User Evaluation Results Google Spreadsheet used to collect data online from partici-

pants who participated in the user evaluation.

google-play-services lib Google Play services APIs library required to allow usage of

Google Play services APIs within a mobile application.

Azure Details Login credentials to the researcher’s Microsoft Azure account and other

details.

78