Upload
vandang
View
223
Download
7
Embed Size (px)
Citation preview
T
CHAPTER 7
RESULTS AND DISCUSSIONS
This chapter summarized the obtained results followed by discussion.
7.1 Obtained Results
his research, exploration is based upon the data gathered from the seven diverse
finished Agile Web based information projects, taken from the computer science
student’s course of major project amid most recent 2 year. Average gathered
information from these Agile Web projects are outlined beneath. It is expected that all the
information of seven distinctive project appeals mulled over is right without any
deviation. For Size computation, we have selected the Agile Web point Analysis. They
are a measure of the extent of web applications along with the activities that construct
them.
Table 7.1: Effort Estimation Results (AgileMOW v/s webMO)
Project ID
Language/ Web Technology Used
Actual Effort Eact [P-M]
Model AgileMOW
Model webMO
AWP Estimated. Effort [Eagm] [p-m]
MRE [%]
Web object
Estimate d Effort [Ewbm] [p-m]
MRE [%]
AP1 JAVA for 7 225 5.23 25.24 154 10.82 54.52 web
AP2 XML 1.9 384 1.20 37.08 202 2.93 54.38 AP3 PERL 21 345 26.76 27.41 266 13.54 35.54 AP4 HTML 11 238 7.65 30.46 201 6.42 41.68 AP5 HTML 11.25 167 15.87 41.06 140 6.69 40.57 AP6 SQL for
Web 13 421 8.60 33.85 255 6.90 46.91
AP7 JAVA for 26.5 268 36.54 37.89 232 45.19 70.51 web
MMRE 33.28 49.15
Using described AGILEM
method, precision of these
illustrated in Table 7.1.
EMOW [Litoriya and Kothari, (2013 B)] Effo
se Web Projects Estimation developed using Agile
Figure 7.1: MRE comparison graph
Figure 7.2: Effort comparison graph
ort Estimation
e methods are
7.2 Discussions
The market fruition forces customers to get included in transient Web projects. Various
Web development organizations worldwide don't utilize formal strategies to gauge effort
for new activities, hence depending on expert-based presumption . Furthermore,
numerous Web organizations don't accumulate any information on past projects, which
can later be utilized to gauge effort for new web ventures, and as a result they are not
mindful of how effort is utilized all through their undertakings and on the off chance that
it could be utilized all the more adequately. This research study discusses about a
parametric model of web cost estimation, we accept can be of profit to Web organizations
to help them enhance their effort estimation rehearses. The suggested and improved
model are especially focused at small Web development organizations. Expert-
based effort estimation speaks to the procedure by which effort for a fresh undertaking
to be created is evaluated by subjective means, and is regularly focused around past
experience from creating or overseeing comparable activities. This is by far the most used
technique for Web effort estimation. Assessments can be proposed by a task supervisor or
by a gathering of individuals blending project managers and developers, normally by
method for a meeting to generate new ideas. In the context of web development, our
experience proposes that pure expert based effort estimates are gotten utilizing different
mechanism.
The results are unswerving and hard to keep up applications that unsuccessful to help,
and cost more and take more time to create than anticipated. By and large, the effort
estimation for a particular scenario is not fit to predict and help dodge these issues. The
results [Litoriya and Kothari (2013,B)] show that the proposed methodology gives an
effective mechanism to measure the extent of software systems, differentiate web
programming frameworks, and estimate development effort ahead of schedule in the web
product life cycle to inside +/ -33 % over a scope of web application. Interestingly, with
other accessible strategy, AGILEMOW utilizes crude recorded data about development
capacity and high granularity data about the web framework to be produced and Agile
group, keeping in mind the end goal to complete such estimations AGILEMOW reduces
the MMRE form 49.15 % to 33.28% which reflects the considerable improvement for
web estimation particularly in context with Agile paradigm. Figure 7.1 and 7.2 reflects
the comparative analysis of MRE and efforts respectively. This technique is
straightforward and exceptionally cited in favor of little or average Web based projects
developed under the agile umbrella.
This dissertation offers results on the significance of the Agile attributes and
model parameters for web project data set developed by undergraduate students of the
computer science discipline. The results are considerable, repeatable and can be
improved by any other researchers. This approach involves identification of crucial Agile
factors (cost drivers) and imparting them in the parametric cost model. Through this
approach, the prediction accuracy of the web based application developed using Agile
methods can be significantly improved while the variance is diminished for the dataset
being analyzed. The approach can perk up the confidence of Agile estimator, demonstrate
the prophetic power of web cost models subsequent to local calibration, and make it
uncomplicated to specialize the models to meticulous situations.
S
CHAPTER 8
SUMMERY AND CONCLUSIONS
8.1 Summery
oftware Engineering is the well defined process which lays the roadmap for
development of software’s within the specified schedule and effort and with the
preferred quality. Software designing is concerned with building programming
concentrated frameworks and items inside the imperatives of time, assets, innovation,
quality, and business contemplations. Presently, software’s are the pouring force behind
the majority of the day to day requirements and service delivery, such as entertainment,
health facilities education, business, transport etc. Each one of these areas update or
maintain technologies that offer quality services to their customers. Nearly all of these
software technologies are expensive, intricate, and entail accurate planning to be
developed reasonably fast. The need for precise prediction of the cost and effort of
software projects are uniformly growing. The capacity to degree ventures faultlessly is a
fundamental component of a software engineering discipline. It entails utmost level of
analyses, hard work and the management of the two. Big scale software planning and
control incorporates the usage of quantitative programming evaluation and estimation
models that are situated in the lead of theory and collected historical project data. With
the perpetually expanding size and unpredictability of software's; software development
has turned out to be a more rowdy process and hence desired to take care of even the
simplest action in the process of software development.
The process launches with assessing the approximate size, required effort and
tentative time required for the maturity of the software and stops with the product and
additional work products constructed in different phases of development. The troubles
being appeared in the software developments is cost overrun, schedule overrun and
quality deprivation. Wrong estimation indubitably results a failure in the development
process. Efficient estimation is essential and important in favor of appropriate project
-
23023
planning as well as for process control and is one of the basic and key testing assignments
in the software development process.
The broad estimation process demonstrated in Figure 8.1 remnants impervious while
applied to either conventional software estimation or Web effort estimation.
Nevertheless, in spite of sharing alike estimation processes, Web effort estimation
diverges from traditional software cousin for a number of reasons :
Figure 8.1 : Broad Estimation process [Mendes, (2007)]
Therefore, a standout amongst the hugest points of the software building expertise has
been to create useful and down to earth models that productively clarify the software
progression life-cycle and exactly expect the effort required and timetable of creating a
software venture.
-
23123
Taxonomy of Existing cost estimation models
Noteworthy research on software cost estimation models begin in the late 60’s.
Subsequently, more robust cost models were developed, i.e. SEER, ESTIMACS,
COCOMO and COCOMO II etc. The literature reports a vast variety of categorization of
the Software Estimation models/methods and techniques. A detailed sorting of cost
estimation methods by dividing them into 6 sorts i.e. A theory based methods, machine
learning methods, empirical methods, regression methods, composite methods and
expertise based methods are presented by a researcher. A categorization given byy [Singh
et al., (2008)] , includes another class the dynamics based procedures. This additional
class of estimation method stress upon the dynamic characteristics of the software
development effort information. A common grouping of effort estimation is proposed by
[Shepperd et al., (1996)] and [Li et al., (2006)]. They grouped them into algorithmic
methods, analogy based, expert judgment, and machine learning.
[Laird and Brennan, (2006)] supplement the various proposed classification of taxonomy
by including strategies utilizing benchmark information, custom models and proxy
points. Models making utilization of a benchmark data allow associations to many-sided
effort estimation focused around existing data acquired by alternate organizations. A
more generalized classification is published by [Leung and Fan (2002)] and [Attarzadeh
and Owe, (2009)] by isolating them into algorithmic and non algorithmic ones. The
whole characterizations referred to above have a typical set of methods, which wording
may change from characterization to grouping, then again, the meaning keeps up the
same. Principally these models were supposed to address conventional software and
development scenario. Though the majority of these models was shaped at about the
same duration, they all experienced the software crisis of increasing size and complexity,
made it hard for the software project team to forecast the cost, schedule accurately. The
ever active area of software cost estimation proceeded with the interests of analysts and
experts who accomplished something in setting the benchmark of software building cost
model. Further, few categories are particularizations of the more basic classifications, this
chapter concentrates on the association of officially existing strategies into Six significant
-
23223
d
classes, including the Web estimation techniques as revealed in figure 8.2, giving a
layout cases of every class.
Model-Based
Learning- Oriented
Software Estimation Techniques
Expertise-Base
Regression- Based
Composite -
Bayesian Web Estimation
Figure 8.2 : Taxonomy of Existing Cost Estimation Models
Over late decades, while market strengths, frameworks prerequisites, usage innovation,
and project team were shifting at a relentlessly expanding rate,
a distinctive program
development approach demonstrated to its favorable circumstances above the
conventional one.
Agile Paradigm
The Agile approach of advancement specifically attends to the issues of quick change. An
overwhelming thought in Agile advancement is that, the players might be more viable in
reacting to change in
the event that it can reduce the expense
of data flow among
-
23323
individuals, and decrease the snuck past time between settling on a decision to
considering the results of that decision.
In view of the fact that software specialists are not machines. They exhibit incredible
variety in meeting expectations, styles; inventiveness, consistency tidiness, the Agile
methods have become an admired topic in software development communities, giving
rise to much mystification and argument. The Agile Manifesto was printed in 2001, and
offers a good quality, decisive point for examining the Agile methods and understanding
the foundation on which they were built [Beck, (2001)]. The four Agile menifestoes are
shown in the figure 8.3 below.
Figure 8.3: Manifestoes (Policies) of Agile development
The Agile Alliance characterizes 12 agility principles which flesh out the ideas
articulated in the Manifesto. These 12 principles were documented during the February
2001 summit. They present a much more concrete view of the type of activities that
comprise the various Agile software development method.
-
23423
The importance of Human aspect
Promoters of agile software development clearly state their confidence that
Human/People are of greater importance informative software project success instead of
processes or tools. Agile development focuses on the talents and skills of individuals,
molding the process to specific people and teams. The critical tip in this announcement
is that the methodology molds to the necessities of the individuals and the group, not the
other path around. On the off chance that software development team allies are to coerce
the attributes of the progression which is helpful to assemble programming, a set of
different imperative qualities must present amidst the individuals on an agile group and
the group itself.
The Agilest are not only advocates, the people. Human or people have been
within the software industry for decades. The software engineering community believes
that if one compose the team of the right people, and attends to their needs, then
individuals will be capable to succeed no matter what. Some say that disciplined
processes and structured tools can also get in the way of project success. Figure 8.4
illustrates the three interconnected aspect i.e. Human, Process and tools for project’s
success. To make one aspect week or underestimated than the other two would make the
project unstable and eliminating any of these would cause the project to fail.
Human being are considered to be most appreciated resource. Industries that create
intellectual property, people have a superlative role, as observed in figure. Process cannot
imagine and create. Tools cannot apply intelligence. Converting ideas into running
software requires human resource. People envision, people understand, people visualize,
people imagine what they suppose to build, and then people transform that vision into
actuality. Writing sophisticated and quality software cannot be imagined without people.
-
23523
r
Human
ool Process
Figure 8.4: Role of Human Aspect in Agile
In 2002, agile ventures made up short of what 2% of general projects and short of what
5% of new application advancement ventures. Today, agile undertakings represent very
nearly 9% of all activities and 29% of new application advancement projects, for a 22%
CAGR. The build in task achievement rates can specifically attach again to undertakings
determined through the agile methodology. The Agile approach is well-liked and admired
by its practitioners and has normally been flourishing when deployed for appropriate
projects. Even though Agile methods commenced with small projects, efforts are being
made to extend the approach to bigger application. As per the p inciples and nature of
Agile approach its applicability is most suited to web based
projects. Web based
applications are relatively small in size and perhaps the most interesting and swiftly
changing field in the era of software engineering.
-
23623
The majority of practitioners who follows Agile practices have communicated
apprehension over their powerlessness to exactly gauge expenses belongs to Agile web
project advancement. This worry has gotten to be significantly more basic as expenses
connected with improvement keep on increasing.
WebApps – An Emergent Picture
Web-based software can be expressed as any application software that is coded in a
browser-supported programming language and dependent on a pervasive web browser
to make the execution of application [Coward, (2000)].
Web based applications are the eventually means to take benefit of modern technology to
enhance any organization’s productivity & efficiency. Web based application provide
business professionals a prospect to access their business information from anywhere at
any time across the globe. It also makes easy to save time and money and enhance the
interactivity with their customers and partners. Web based applications generally make
use of an arrangement of client side script such as JavaScript, HTML etc. And
server side script, for instance ASP , PHP etc. to build the application. The appearance
and arrangement of the information is managed by client side script, whereas the server-
side script manages all the complex tasks like storage and retrieval of the information.
Web applications are well-liked owing to the universality of web programs, and the
handiness of utilizing a program as a web client.
The amount of web locales on the worldwide internet was near about 634 millions in
December 2012 [Pingdom, (2013)]. This figure simply confirms an impression about the
tremendousness of information open all through the Web. The Web is continuously used
as the conveyance stage for adequate sorts of uses, fluctuating from on line commerce (e-
commerce) application to backend databases to individual static Web pages. A key
reason behind the attractiveness of web application is the potential to upgrade and sustain
these applications devoid of sharing out and installing supporting software on probably
numerous customer machines. It provides the inbuilt support for cross-platform
companionability. Normal web apps incorporate online retail deals, wikis, webmail,
-
23723
online barters, and web portals in addition to numerous different functions. The latest
push for web applications is replacing the conventional web applications; these don't
typically require a server to store the data. For example word processor, it stores reports
on our machine, and doesn't need a server. Web applications are capable of providing the
same usefulness and increase the advantage of operating transversely multiple platforms.
Web Cost Estimation
The field of Web effort estimation is relatively new, particularly the estimation area in
Agile development is very recent, it would not be surprising that research findings in
this field may not have reached an industry widely. A foundation stone of Web based
software project management is sound to cost estimation, which is the procedure all the
way through which an expert, a group of experts, cost model, tool, or a grouping of these,
envisages the cost involved in a given task.
In the context of Web software development, cost estimation, speaks to the
achievement of an expected effort for a fresh web project to be produced. Such an
effort estimate can be gotten by expert opinion and/or the utilization of unequivocal
models, those are explained in chapter 3 of this thesis, constructed from past task
information. The major challenge here is to acquire reasonable effort estimates, that is,
guesstimates that are close in quality to the actual effort it will take to finish the new
extend. Right now, there are various challenges in the matter of why effort assessments
need precision. The communities of software cost estimation presently has not conceded
to how to create effectual estimates for Web based software projects. The inconvenience
is that the attributes of the Web based tasks make it troublesome for estimation team to
adjust and put existing methods and models. In the light of various challenges and
difficulties Web cost estimation becomes very significant particularly for the newer
development strategies.
-
23823
Web based software development are undoubtedly not the same as ordinary software
improves and, thusly, involve diversely customized routines for impeccable effort
estimation. In the specific environment of Web based development, these concerns are
likewise fundamental and extremely urgent, realized that Web based tasks have petite
calendars and an exceptionally fluidic extension, researchers and professionals have
anticipated and compared cost estimation procedures for Web based ventures utilizing
Software Company and academic datasets over the last 15 years.
Enhancing the quality of web effort estimation is a prime necessity in numerous web
development firms. A consistent quest is always there for improved models and
instruments to help task administrators in their assessing procedure, in specific for the
Web app market, which is picking up a noteworthy offer of the aggregate IT market.
AGILEMOW : A Regimented approach to Agile-web effort estimation
This thesis discussed the necessity for new measurements and model to fast predict the
size and effort for agile based web development projects. The method called web
estimation using COCOMOII for Agile methods AGILEMOW, attends to a need to
obtain web effort estimation in a restricted period using human oriented cost factor
information. Conversing with other existing techniques, AGILEMOW uses agile
characteristics, crudely recorded data concerning development capability along with high
granularity knowledge about the web framework to be created.
Prescient methods concentrate on the design of a customized plan to do the project.
Conversely, versatile methods concentrate on inspecting the current circumstance and
discovering the best result in every step of performing the task [Azadani and Doost,
(2008)]. The initial step of constructing AGILEMOW (Cost estimation of Agile Web
based software) is focused around a compound web object and various traits (attribute)
approaches.
-
23923
The research study carried out in this thesis, focuses on development of effort estimation
model for agile web projects. Creation and use of the model is enlightened in detail. The
model was balanced utilizing the exact information gathered from scholarly developed
seven completed web projects.
Sizing Web: The Agile Web Points (AWP)
AWPs like other aberrant measurements such as FPs, Object Points or Web Objects
which indicates users concerning the comparative extent of an application in the light of
truth that they symbolize dynamic ideas that are used to get the span of the web
framework to be developed. In meticulous, AWPs similar to Web objects signifies system
functionality from the perspective of its web entity predictor (Web entity predictors are
the, visual, aural or text content that comes across as part of the user experience on
websites. It may contain, among other things like text, images, sounds, videos and
animations).
A key topic in the origin of Agile development model is the need to improvingly
comprehend, assess and handle change while Agile software is planned and built. It is a
genuine truth that requirements will definitely undergo change [Fowler, (2005)]. Clients
of web products will alter their opinion, their impression of the part of the product will
change, nature's domain in which the product works will change along these lines will the
engineering with which the software is built.
Without the inclusion of the agility factor while computing precise size metric for agile
web development is iniquitous. Agility factor (AF) for agile web size estimation is
defined, as a gauge metric calculated for each web project for the purpose to determine
the agility of particular or group of web point.
Therefore, agile experts analyzing the web entity data obtained through specification will
get the exact size for a framework; however the development effort might be contrasted
relying upon specific qualities of the agile advancement environment, for example,
-
24024
people centric features like the aptitude of the development group, and the accessible
supporting instruments, among others. AWP expects to make things less demanding the
effort estimation transform by giving the expert better devices for adjusting this
methodology. This AWP measuring metric has the point of interest of being quick and
simple to apply, every time the obliged chronicled data is accessible. This makes it
particularly suitable to estimate Web-based projects in agile environment.
AWP Table computes the fairly accurate size size of web based system. The weight
allocated, i.e. Low, medium or high to every classification of AWP means the web
advancement effort of each one, and it is focused around the learning and practice of
the expert agile estimator. The weights may be adjusted by the agile estimator, each time
he accepts it is right.
Agile Web Effort Estimation
The AWP size metric is an approximation of the whole volume of the web project and is
simply the initial step towards creating a model for precisely predict effort of Web
applications developed using agile methods. To precisely estimate agile web project
costs, an AGILEMOW cost model has developed and suggested. It is affirmed, the
suggested model is an adaptation and extension of the COCOMO_II [Boehm et al.,
(2000)] and WEBMO [Reifer, (2000)] model. The0proposed model is created utilizing a
mixture of professional opinion and genuine information from completed academic
projects developed by computer science students of undergraduate studies. It facilitates
the agile team to capture the characteristics of the web application as well as Agile
methods efficiently. AGILEMOW’s algebraic formulation builds on the COCOMO II
model with a far reaching information investigation of web projects and critical
environmental condition are considered to address agile Web estimation issues. Agile
environmental factors can have a major influence over the development efficiency
of the development team.
-
24124
bc
TUVWX YXZ L[[\]^ (L) = T _ ̀aYTV (TYf)fe
Vde
Where, A is the constant
P1 is the power factor
CFWAi are each of the Cost Factors of Agile Web development.
AWP is size in KLOC
The 23 cost factors (CFWAi ) consequently effort multipliers of the Suggested
AGILEMOW Architectural model are described in chapter 6, few of them are analogous
to those described in WEBMO [Reifer, (2000)] and remaining all are derived from the
environmental condition of agile strategies. All these twenty three variables are
calculated qualitatively by choosing a rating from a well-defined rating scale. The last
adjustment constant in AGILEMOW corresponds to power factor P1 namely the value of
the exponent of the AWP. The value of this exponent is close to 1.00, and it should not,
one or the other is higher than 1.10 or lower than 0.99. In the AGILEMOW effort
equation the constants and power laws for all of the 3 application domains are directly
taken from WEBMO [Reifer, (2000)].
Experimental analysis and model caliberation
The experimental results prove that the model has good estimation accuracy in terms of
effort and MMRE. The suggested method/model is straightforward and especially proper
for little or medium size web frameworks created by making the utilization of Agile
procedures. The formulation of AGILEMOW is elaborated in chapter 6. Calibration and
-
24224
Validation of the metric and detailed model is done in expectancy of developing potential
Agile Web products based upon them. A starting alignment for AGILEMOW by
consolidating expert judgment and real information from 7 finished Agile based
Web projects is performed. The ultimate goal is to provide accurate estimates to the
Agile practitioners who use to develop web based application. Using descriptive
AGILEMOW Effort Estimation method, exactness of these Web Project Estimation
created utilizing Agile methods are captured. The results show that the proposed
methodology gives a productive instrument to measure the extent of software systems,
differentiate web programming frameworks, furthermore evaluate the development effort
early in the product life cycle to inside +/ -33 % over a scope of web application.
8.2 Conclusions
The past works on Software cost (Size, Effort) estimation has concentrated on generating
the cost estimation model for general software development paradigm from the available
data. With the advent of relatively newer web development scenario, considerable
research has carried out (chapter 5). When Fast and flexible Agile methods (chapter 4)
came into the existence and widely adopted by Software practitioners, a tailored version
of existing and established estimation method is highly desirable, that can effectively and
accurately address size and effort estimation issues of Agile web development. In this
exploration, the parametric software cost estimation models (Chapter 2) and the allied
calibration methods have been investigated, particularly for the COCOMO [Boehm et
al., (2000)] and WEBMO [Reifer, (2000)] calibrated approach. It is found that these
models have used some limited parameter/cost driver/cost factors that do not necessarily
holds the characteristics and attributes of Agile methodologies to prove the accuracy of
the model. The best and suitable web cost estimation model is found to be WEBMO
[Reifer, (2000)] with 9 varying cost drivers. The results of this parametric model are less
accurate and shows inconsistency when applied for web application projects developed
using people centric Agile methodologies. The major contributions of this dissertation are
as follows.
-
24324
1. This thesis introduces an outline of a mixture of programming, estimation
techniques, giving a sketch of a few prominent estimation models presently
accessible. The literature to date shows that AI based procedures are less
experienced than parametric modeling techniques; however, that majority of
methods is faced with the quick pace of progress in software innovation.
2. The AGILEMOW approach discussed in chapter 6 addresses one of the
significant issues confronted by the Agile web building society; the challenge of
making great and solid estimates utilizing data that is generally inadequate and
deficient. The majority of the contemporary empirical software engineering
cost models is aligned utilizing historical data that do not fit for modern practices
and the environment. The chapter presented the development of the new size
metric AWP based on Web Object. This tailored size metric includes Agility
factor in its formulation so as to effectively measure the Agile web project size.
3. On request to secure quick and solid effort estimations of Web
based information framework improvement projects developed using a
specialized Agile scenario, a web effort estimation model for Agile Software
projects has been presented. Considering different challenges faced by the agilest,
the model is proposed and validated to accommodate most of the characteristics
of Agile methodology, especially people centric attributes. The proposed model
that consists of 23 cost factors that surely influences the development cost is
practically implantable and does not supplant the master estimator, however, it
furnishes Agile development group with a device for accomplishing a more
precise estimation, taking into account relevant information in a shorter time. The
work discussed in this thesis is based on the original WEBMO model proposed by
Reifer.
-
24424
4. An all-inclusive framework for calibrating the suggested model employing the
Delphi Expert judgment and Bayesian methodology to create robust web cost
estimation models was exhibited in the thesis. This approach has worked
effectively on numerous models like COCOMO II and its other extensions and is
being employed for AGILEMOW. The framework endows with a legitimately
unswerving and prescribed way of making use of practice based expert
judgment data alongside sample information in the decision making procedure.
5. An efficient, cost estimation model for Agile web development, extension to
COCOMO II and WEBMO has been created. Owing to the absence of historical
data of relevant project to empirically calibrate AGILEMOW, a Delphi decided
a-priori model has been produced and further calibrated using a Bayesian
approach. The model portrayed in chapter 6 can be utilized in its present
condition otherwise can be locally attuned to a specific organization anticipate the
size and required effort of the web based software being developed on.
8.3 Recommendations
The answer for enhancing estimation exactitude for diverse development technologies is
not a high innovation issue. No existing devices, models, or philosophies might be
presented as a powerful influence for the issue that independent from anyone else will
have a huge effect. Few recommendations based on this work are as follows
1. The cost estimation issue changes extensively among software development
strategies that do their estimation under altogether different environment
conditions. The recommendations are specific in nature and must be applied to the
web based software project developed using Agile methods.
-
24524
2. Web based projects are entirely different from the conventional software project
types. They require a specialized estimation model or tool for predicting accurate
cost.
3. Agile software development is again different from conventional development
methodology. The most paramount ramifications to organizations working in the
Agile way is that it puts more attention to the characteristics and qualities of
individuals considers in the project. These qualities turn into an essential concern
for the would-be Agile group. Therefore estimators should prominently consider
this “human” factor during estimation of such projects.
4. Agile practitioners and Designers can utilize this new web measuring tool called
Agile Web Point and an effort model which is an adjustment of the WEBMO and
COCOMO II model called AGILEMOW to more effectively estimate the
development effort of Agile Web based software projects to be developed. In view
of research work with in excess of 7 scholar projects, suggested tailored version of
estimation model is particularly helpful for small and medium sized web projects.
8.4 Limitations
Though the model is an extension and elaboration of COCOMO II and WEBMO, most of
the preconditions are same. But due to the fact that it is de-signed specifically for the
projects developed within a flexible and rapid environment, the prerequisite condition or
limitations of this suggested model to be functional are listed below
1. It is explicitly designed for web projects so the project supposed to be estimated must
be a web application.
2. It addresses the estimation issues of Agile development environment, therefore the
development environment of the web project must be any of the Agile approach.
-
24624
3. Because of the fact that this is the introductory model and calibrated and tested over
few academic projects, therefore the approximate size and scope must not exceed
medium range for a web application.
4. The qualitative data collected by the development team must be factual.
5. It is not necessary that the historical estimates collected are from the same team. The
different web project may have different team of variable size and attributes.
6. Since the Agile activities and its allied methodologies came into the picture after
2001, therefore the project data must not be older than 10 years.
8.5 Future Research Scope and Extension of Work
1. WEBMO has nine cost drivers and this makes it complex to statistically adjust or
calibrate the model without running into problems of over fitting. Though,
the argument is that WEBMO is a complete web cost estimation model eliciting the
entire web development life-cycle. Now consider our data gathering process, when
we faced many problems with the selection of parameter (cost factors), Develop for
(EORG), giving a strong indication to drop the variable. In actual fact, eradicating
a predictor variable is equivalent to stipulate that variations in this variable,
have no consequence on project effort. When the professional in the field and
the detailed behavioral analyses recommend otherwise, very strong evidence
is needed to drop a predictor. For this reason, further research wants to be done to
resolve this issue either by gathering more data or by accepting if fewer
than twenty three parameters can be sufficient to develop a more cost-conscious
model without giving up its coverage.
-
24724
2. Given the restricted data points available, AGILEMOW confirms good expressive
value in modeling the relationship between various cost drivers and required project
effort for agile-based web projects. The variation of sub-factor values within used
dataset is restricted, however, due to the environment of the Agile web development
firm performing the work and the domain of projects it executed. It is inexplicable
whether the model will perfectly describe the estimates outside this limited set. On
further examination, we expect to collect further more data points over a wider
assortment of Agile web projects. This data is supposed to allow us to better
calibrate the model and assess its wider applicability.
3. Any parametric cost model that doesn’t use historical data has faced the problems of
measurement error (errors-in-variables problem). This is a contravention of the
assumption made by the multiple regression approach and the Bayesian approach.
Hence, for the intentions of this dissertation, the supposition has implicitly
been made that the random variation in the responses of the various parameters,
little contrasted with the scope of the parameter. But, there exists enough
literature on research done in other provinces that attempt at minimizing the effects
of measurement error. A study is recommended to comprehend how these problems
can be resolved to enhance the accuracy of the AGILEMOW and other models
would be highly recommended.
4. The A-priori values for the calibration by Bayesian framework is acquired
by sampling experts and researchers in the field of study and receiving their view
on impact of parameters on the measures to be estimated. It would be attention-
grabbing that how the prior information matures and how it is supposed to be
utilized for future calibrations when rationalized sampling data is available.
5. The present proposal of AGILEMOW is limited to compute effort for the entire
web development life cycle. A future research activity may possibly
involve collecting data on effort depleted in various activities and studying the
division of effort by activity.
-
24824
6. The dissertation presented improvement in accuracy when the broad development
scenario is Agile and software to be developed is web based. A similar study and
development of customized, cost models according to the varying methods of Agile
paradigm as discussed in chapter 4 would be very interesting.
7. Creation of a dynamic model is desirable that enables critical path analysis and
updating. This type of model could be used to update estimates while more data
become obtainable with the development progress.
8. The current results of AGILEMOW model don’t illustrate any significant
differences in prediction accuracy among small and medium size web projects. It
would be motivated to do an analogous study to see if there is an impact on accuracy if
the size increases from medium to large. Such a study can be made after the
collection of sufficient and quality data.
Software estimation researchers and model developers, continue to face the hindrance of not
existing any universal laws of “Software mathematics ” that delineate quantitative relationships
among the variety of independent variables describing the product and the environmental
aspects of the project. Even though some approximate solutions can be found.