14
Location Analytics for Optimal Business Retail Site Selection Ahmad Murad Bin Mohamed Rohani and Fang-Fang Chua (&) Faculty of Computing and Informatics, Multimedia University, 63000 Cyberjaya, Selangor, Malaysia [email protected], [email protected] Abstract. The issue on location placement for next business establishment is always a challenging topic. It presents businesses with many opportunities to uncover the most sophisticated approach on selecting the next location of physical stores to establish its presence. The traditional approach of manual survey of land, competition landscape and also related to demographic factor analysis comes with high cost and longer time to complete. Our proposed work leveraging Google Maps to survey the surrounding and records the existing characteristics such as whether the shop is a corner shop lot, can be viewed from main road or having a sizable parking space. Based on the ndings, the char- acteristics listing mainly relates to the business type. The approach of this paper can be used as one of the alternative input to decision making for physical placement of store. With the proposed work, optimal store location placement is determined based on a set of characteristics of an existing location. This research may help new business to gain optimal in ux of customers based on the location identied. Keywords: Location analytics Association rule mining Retails placement 1 Introduction Much effort need to be considered from inception till materialized for investing in a new place for a brick and mortar kind of establishment. Factors affecting the locational decision making includes external & internal environment, locational management activities and their portfolios [1]. For instance: Where will your next shop be located? Is it near the main road or a corner shop lot? What are the important aspects in the surrounding area that need to be observed in order to assure that you have the best return from the investment for the new place? These are part of the questions that need to be considered when choosing the location for new shops. Due to its xed in nature, location cannot be modied in a short period of time and this is conicting with elements such as price, customer service, or advertising [2] which can be adjusted accordingly following the needs and demand using forecasting and other related tools. This research will leverage the availability of data over the internet via web medium by examining location characteristics and the corresponding business types that exist within the nearby area. Then, data mining techniques are performed to determine the top ranking characteristics that exist based on a pool of recorded © Springer International Publishing AG, part of Springer Nature 2018 O. Gervasi et al. (Eds.): ICCSA 2018, LNCS 10960, pp. 392405, 2018. https://doi.org/10.1007/978-3-319-95162-1_27

Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

Location Analytics for Optimal Business RetailSite Selection

Ahmad Murad Bin Mohamed Rohani and Fang-Fang Chua(&)

Faculty of Computing and Informatics, Multimedia University, 63000 Cyberjaya,Selangor, Malaysia

[email protected], [email protected]

Abstract. The issue on location placement for next business establishment isalways a challenging topic. It presents businesses with many opportunities touncover the most sophisticated approach on selecting the next location ofphysical stores to establish its presence. The traditional approach of manualsurvey of land, competition landscape and also related to demographic factoranalysis comes with high cost and longer time to complete. Our proposed workleveraging Google Maps to survey the surrounding and records the existingcharacteristics such as whether the shop is a corner shop lot, can be viewed frommain road or having a sizable parking space. Based on the findings, the char-acteristics listing mainly relates to the business type. The approach of this papercan be used as one of the alternative input to decision making for physicalplacement of store. With the proposed work, optimal store location placement isdetermined based on a set of characteristics of an existing location. This researchmay help new business to gain optimal in flux of customers based on thelocation identified.

Keywords: Location analytics � Association rule mining � Retails placement

1 Introduction

Much effort need to be considered from inception till materialized for investing in anew place for a brick and mortar kind of establishment. Factors affecting the locationaldecision making includes external & internal environment, locational managementactivities and their portfolios [1]. For instance: Where will your next shop be located?Is it near the main road or a corner shop lot? What are the important aspects in thesurrounding area that need to be observed in order to assure that you have the bestreturn from the investment for the new place? These are part of the questions that needto be considered when choosing the location for new shops. Due to its fixed in nature,location cannot be modified in a short period of time and this is conflicting withelements such as price, customer service, or advertising [2] which can be adjustedaccordingly following the needs and demand using forecasting and other related tools.

This research will leverage the availability of data over the internet via webmedium by examining location characteristics and the corresponding business typesthat exist within the nearby area. Then, data mining techniques are performed todetermine the top ranking characteristics that exist based on a pool of recorded

© Springer International Publishing AG, part of Springer Nature 2018O. Gervasi et al. (Eds.): ICCSA 2018, LNCS 10960, pp. 392–405, 2018.https://doi.org/10.1007/978-3-319-95162-1_27

Page 2: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

information from various location establishments. The result later can be aggregatedwith other traditional methods for the decision on the next retail business establishmentinvestment and in turn help to increase the turnover of the business. With the correctinformation and approaches, they can make an intelligent choice for a good place andperhaps will yield the most visitors.

This paper aims to achieve the objectives based on the data mining techniques andanalysis: (a) to propose the attributes of location characteristics that influence theplacement of retails shop. (b) to analyse the characteristics of location placement andpropose the method for selection. (c) to evaluate the proposed method by using the datamining techniques. Relationship between the characteristics of a location with a par-ticular business is extracted. Examples of businesses type include mini market, foodand beverages, laundry, etc. Using Google Maps, this study surveys the area of interestand records the attributes. These attributes can be in terms of characteristics of locationand type of stores within the surveyed area. Beginning with selecting the location as thestarting points of search, this research will drill down to the particular site and startrecording the attributes of the surrounding within predefined radius. Each occurrencewill be recorded as binary of zero (0) for non-existence and one (1) if it exists. Based onthe collected information, mining will be performed to find the association among theattributes and types of stores that typically exist. From the result, suggestions related tothe store and location characteristic will be proposed.

Section 2 provides an overview of related works. In Sect. 3, we describe theresearch methodology that uses Google Maps for data collection and how it is used inthis proposed work. Subsequently this paper will elaborate the analysis of findings inSect. 4. Section 5 presents the results and discussion with some research contributionand finally, Sect. 6 presents the conclusion.

2 Related Works

The previous work on selecting location factors are mainly using the traditionalqualitative approach of interviewing, observation and survey [3]. The newer approa-ches of location analytics are using rich social media data such as Facebook [4] andFoursquare [5]. Another online resource is using the search query data from BaiduMaps [6]. We are focusing on location attributes concerning the physical characteristicsusing association rule [7–9] which produce ranks by using lift as a result for relatedlocation characteristics given a business type.

2.1 Location Characteristics Attributes

Selection of location is always being prioritized in order to start a business. Thelocation characteristics refer to the feature or quality belonging to a place that can beobserved. This is important because the retailers can create differentiation throughlocational preferences based around the projection of image and identity [10]. Thecharacteristics of the surrounding location encourage the gathering behavior andcontribute to place attachment in selected coffee shops. The store must also be rela-tively close to major roads, the ability to walk from surrounding neighborhoods. An

Location Analytics for Optimal Business Retail Site Selection 393

Page 3: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

access to nearby shops, parking availability, along with other exterior or site consid-erations also needs to be considered for shop placement [3].

Other research on location selection is based on the largest number of check-insusing data taken from Foursquare check-ins that published in twitter [5]. This check-indata is used to frame the problem of optimal retail store placement in the context oflocation-based social networks. The study focusing on three factors, which are thedensity, heterogeneity and also competitiveness of a place. Method being used is byproducing ranking based on data mining technique. This includes the NormalizedDiscounted Cumulative Gain (NDGC) and also supervised learning such as SupportVector Regression and Linear Regression. The ranking of the location is using theRankNet which is based on Neural Network algorithm. Information on locationselection also using a different approach through the search query requested to thepages via Baidu Maps [6]. Pair it with the location features, the supply side is alreadyshowing the potential customer for the targeted location.

Research on location analytics using Facebook page data of user check-in andidentify the key feature that correspond to the suitable matric for business popularity[4] is another methodology of using online data. The research employs gradientboosting machine as the predictive model to gain insight on feature importance metricsfor location selection. A set of relevant characteristics extracted and a predictive modelis developed to estimate the popularity of a business location. The research reveals thatthe popularity of neighboring business is the key features to perform accurate pre-diction. The more popular the shops around the area, with more check-in, the moresuitable it is for new business to be established around that area. Another research onlocation strategies indicates that one of the primary motivation by the retailers inplacing their outfit is based on the close proximity to capital city residents, due to theirhigh discretionary incomes which likely to become the customers [11].

2.2 Types of Store

The shop characteristics is an important aspect when choosing the location. Forexample on stores carrying the fashion based merchandize, they require more selectivelocation when placing their establishment [12]. The researches focus on the expansiondirection of fashion designer retailers within central area city of London and New Yorkand investigate the main factor driving the chosen location selection. The new retailplaces, spaces and sites includes the department store and the mall where it confines tocertain location such as the primary business street of towns or cities, especially in theUnited Kingdom [10]. As for the street, it reveals that the retailers choose certainlocation in order to create differentiated spaces of consumption through locationalpreferences, which contributed to the exclusive image and identity. This is importantbecause through differentiation it will create the brand exclusivity and attract morepeople who are rich or of a high social class [13]. Other example of store includes theshoe stores aggregate at the town shopping center, while furniture stores are partiallydispersed on secondary poles and drugstores are strongly dispersed across the wholetown [14]. The criteria for each location are different because it attracts certain type ofcustomers with particular needs.

394 A. M. B. M. Rohani and F.-F. Chua

Page 4: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

2.3 Location Intelligence and the Use of Spatial Data

Businesses use data to improve their company performance. More location-based datacan be collected using applications like Google Maps and Microsoft’s Bing Maps.These applications provide free or low-cost and comes with easily accessible mappingcapabilities. According to [15], it allows people using the simple Application Pro-gramming Interface, to put data on maps and integrate the analytical and geospatialcapabilities. Location Intelligence is the ability to process complex data using BusinessIntelligence and geographic analysis [16]. It positions business and geographicallyreferenced data to uncover the relationship of location to people, transactions, facilities,events and assets. Other research introduced Smart Data Localization to promote thedesign and implementation of a Decision Support Systems [17]. The step involvesinclude data extraction and modelling, integration of business data with spatial data,data analysis and also data visualization.

The research on co-location pattern proposes a notion of user-specified neighbor-hoods in place of transactions to specify groups of items. This define the spatialco-location rule as well as interest measures, and propose an algorithm to findco-location rules. It determines that the problem for patterns discovery for the subset offeature located together is different from the association rule problem [18]. Retailersmake use of the spatial technology to acquire new client, retain the existing/currentcustomers and stay competitive with changing user requirements [19]. Research by [8]on discovery of spatial association rules in geographic information databases proposedan efficient method for mining strong spatial association rules in geographic infor-mation databases. However, for the spatial association, the rule indicating certainassociation relationship among a set of spatial and possibly some non-spatial attributesare not straightforward and can lead to misinterpretation.

3 Research Methodology

This research followed a classic data mining process and depicted as diagram below(Fig. 1).

Fig. 1. Research methodology

Location Analytics for Optimal Business Retail Site Selection 395

Page 5: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

3.1 Data Collection

The study begins with the selection of location as a base target for the data captured.We targeted Klang Valley area in Malaysia where most of the businesses establish theirpresence. A favorable location has the ability to meet users’ demands, having lesscompetition and also can attract a significant number of customers. In the proposedwork, we aim to first figure out what are the characteristics that exist within the vicinityof the retails shops and then produce some ranking based on the statistical analysis. Thelocation selected is using a retail shop as a base target for data capture. We have chosenthe one of the successful Malaysian home grown mini market, 99 Speedmart as thestarting location. This store was chosen for two reasons:

i. This establishment has grown to be the largest mini markets in Malaysia with salesup to RM 2.1 billion in 2014 [20], surpassing more shops such as 7-eleven.

ii. The list of all locations where 99 Speedmart located can be easily accessed throughtheir own website (http://www.99speedmart.com.my/) under store locator. We canchoose the desired location where in our case is the Klang Valley region, whichincludes Kuala Lumpur and Selangor area.

Data gathered through observation using Google Maps based on the listed location.All the observation will be recorded as “1” for having the characteristics and “0”otherwise. From the observation, all the characteristics of the surrounding within theradius of 100 to 200 m from the shop are recorded. This radius is considered as the sizeof retails trade area. This area is normally determined by the types of goods offered atthe retail outlets. For example, in our case, the mini market is selling products whichare easily substituted and affordable by majority of customers. This creates a smallerretail zone compared to exclusive store such as Cars outlet [19]. The radius is alsoconsidered an optimal location because it defines as the distance that a customer iswilling to travel in order to buy the grocery [14] and also how newly open store in thearea that will potentially attract the largest number of visits. The characteristics iden-tified in this research include:

i. Facing road view (visibility from main road)ii. Ample parking spaceiii. Corner shop lotiv. Near other building e.g., school, hospital, etc.…v. Near other type of shops e.g., laundry, restaurant, etc.…vi. Near other type of establishment e.g., clinic, dental, offices, etc.…vii. Near public transport e.g., LRT, MRT, etc.…viii. Near Housing area (related to distance to the potential customer)

Preliminary examination reveals that most of the time the criteria are not fixed fromone location to another. Furthermore, number of different business types are increasingas new location is added and this also increases the number of variables.

396 A. M. B. M. Rohani and F.-F. Chua

Page 6: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

3.2 Data Preparation

Our data values are using quantitative variables belonging to a set of items. In our case,the raw data are originated from the original source, which is using Google Maps and isdifficult to be analyzed in its original format. These data need to be pre-processed andconverted to a processed data before any analysis can be performed. Consistent data isdata that is fit for statistical and predictive model building [21]. Missing values, specialvalues, (obvious) errors and outliers are either removed, corrected or imputed.

3.3 Data Mining

We adopt the association rule using Apriori algorithm and also classification techniqueswith Decision Tree to help in extracting meaningful information from the collected data.

3.4 Proposed Framework

The framework used is adopted from research by [6] on ranking the best location fornew business. The difference is on the target output where the original frameworkfocuses in ranking the location whereas in our research, we are ranking the attributes.

Referring to Fig. 2, in one region, there can be many locations to choose foropening a store. Our first step is to determine several places Ld = {l1, l2, l3,…, lm} ascandidate locations for a store Category C or a specific brand of chain stores B. GoogleMaps data is used to find the specific place where the attributes and the type ofbusinesses within the particular locations will be recorded for further analysis. Thenthese attributes will then be transformed into a ranking problem. Our goal is to producethe ranking list of attributes based on the business selected. We achieve this by usingassociation rule based on lift.

Association Rule Mining – AprioriApriori algorithm is used for spatial association rule. A spatial association rule of theform X ! Y where X and Y are sets of predicates and some of which are spatial ones

Fig. 2. Overall framework

Location Analytics for Optimal Business Retail Site Selection 397

Page 7: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

[8], this rule describes the effect of one or a set of features by another set of features inspatial data set. The reference feature is referring to a selected business area and we areusing the spatial association rule with the desired business type to identify the relatedcorresponding characteristics. The characteristics ranking is based on the outcome ofthe association rule mining using Apriori based on their Lift’s score. In general,association rule mining can be viewed as a two-step process [22]:

a. “Find all frequent itemsets: By definition, each of these itemsets will occur at leastas frequently as a predetermined by minimum support count.”

b. “Generate strong association rules from the frequent item sets: By definition, theserules must satisfy minimum support and minimum confidence.”

Classification TechniqueClassification is used to predict group membership for data instance [23]. The aim is toexamine what Business Type that is likely to appear when we set the characteristics ofthe location. Decision tree is used to predict the type of businesses that will likely tooccur based on the chosen characteristics. This is important in order for us to build theprototype to predict the Business type based on location characteristic chosen. We traina supervised ML with the attributes of a particular location. The features are minedfrom an area Ld. The location centered at Ld with radius r = 200 m as this yields thebest experimental results [5].

4 Analysis of Findings

4.1 Data Overview – Descriptive

From the pre-processed data, we proceed with the processing and generate an initialoverview of the data based on the data summary.

Table 1. Descriptive data for three attributes

District

Petaling Jaya : 221Puchong : 118Shah Alam : 93Subang Jaya : 78Seri Kembangan : 57(Other) : 165Total = 20 level

(a) (b)

Business Type

EateryPlace : 125Mini Market : 45Saloon: 39Clinic : 37Learning : 35(Other) : 451Total = 58 level

Business Name

99 Speedmart : 457 Eleven : 12Hong Leong Bank : 4KK Supermart : 3POS Malaysia : 3(Other) : 665Total = 662 level

(c)

398 A. M. B. M. Rohani and F.-F. Chua

Page 8: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

We separate the table based on the attributes with multiple levels from Table 1. InTable 1(a), the District Petaling Jaya has the highest number of location count. Thespecific location is in various places. This district is considered as one of the mostcritical business area with high density of population. Table 1(b) showing the highestcount for Business Type in our data set. The Eatery Place is dominating the count of127 outlets. Corresponding outlet under this category includes restaurant, café andother related food establishment found in the area of observation. We have chosenlocation characteristics as listed in Table 2.

4.2 Data Integration – Correlation Analysis Using Fisher’s Test

Correlation analysis is to remove data redundancy in a dataset. This analysis measuresthe strength of association between two variables and the direction of the relation-ship. Our data consists of multi variables and we would like to measure how theDependent Variable = Business. Type is significant with each variable. Beginning withthe null hypothesis (Ho) where each variable is independent from each other with thep-value is >0.05. However, if the p-value < 0.05, the Ho will be rejected and thealternative hypothesis (H1) will be accepted, where the variable is significant with thedependent variable. To proceed further, we run the Fisher Test on each attributes.Tables 3 and 4 shows the result separated between p-value < 0.05 (statistically sig-nificant) and p-value > 0.05 (statistically insignificant):

Table 2. Location characteristics with details description

Locationcharacteristics

Description

Ample parking The surrounding has a sufficient parking place. For example, availabilityof parking on each side of the road in front of the stores or near toparking area

Corner shop lot The shops are located at the end corner shop lotHousing Near to housing area such as an Apartment or a common

accommodation are such as Taman or KampungFacing road view The stores are visible from the outer road or main roadRoad entry Entry to the stores is located near to road entranceHospital Near to Hospital such as government HospitalLRT Near to LRT or public transportSchool Near to School such as Primary or Secondary government schools

#No of location characteristics = 8

Location Analytics for Optimal Business Retail Site Selection 399

Page 9: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

4.3 Mining Process

Association RuleWe are interested in the top Businesses that appear in our listing in Table 1(b). We setthe right-hand-side of the Apriori function following the chosen business type ofinterest. The support and confidence percentage will be changed accordingly following

Table 3. Attributes with p-value < 0.05

# Attributes p-value # Attributes p-value

1 CornerShopLot 0.0004998 15 Hardware 0.0064972 FacingRoadView 0.04948 16 Hotel 0.046483 RoadEntry 0.005997 17 KitchenCabinet 0.0024994 24HrsConvenience 0.0009995 18 Laundry 0.00099955 BakeryShop 0.007996 19 Learning 0.00049986 Bank 0.0004998 20 MedicalShop 0.012497 CarAccessories 0.01399 21 MotorcyleShop 0.00049988 CarWorkshop 0.0004998 22 Office 0.0024999 Clinic 0.0009995 23 Pets 0.0169910 EateryPlaceMamak 0.004498 24 Telecommunication 0.00199911 EateryPlace 0.002499 25 Travel 0.000499812 Saloon 0.0004998 26 WellnessCentre 0.00699713 FashionAndClothing 0.0004998 27 WorkAgency 0.00349814 FitnessCentre 0.01299

Table 4. Attributes with p-value > 0.05

# Attributes p-value1 AmpleParking 0.303302 Housing 0.389803 Hospital* Error4 LRT 0.494305 School 0.258406 Bookshop 0.568707 OtherMiniMarket 0.740608 FastFood 0.203909 FruitsShop 0.4058010 Music 0.1119011 Optical 0.0729612 Pajak 0.0894613 Printing 0.1384

*x having only 1 value

400 A. M. B. M. Rohani and F.-F. Chua

Page 10: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

the number of rules outcome that is suitable to be used. This is based on theright-hand-side parameter following the Business Type. The ranking of the attributes issummarized in Table 5:

Classification – Decision TreeWe proceed further with the supervised learning using classification method. This is topredict the Business Type based on the location attributes. The rationale of usingdecision tree is to get the insight from the data based on the known set of data. Theinsight is related to the Business Type that likely to appear based on the surroundinglocation characteristics. From the information on the ranks of attributes, we can makethis as the predictor to determine the possible Business Type that might be suitable to

Table 5. Ranking of attributes based on the Business Type

Business Type Attributes Ranking

Eatery

Place

1. CornerShopLot=yes,2. RoadEntry=no, 3. Clinic=yes,4. Office=yes,5. Pets=no6. Saloon=yes, 7. CarWorkshop=yes, 8. WellnessCentre=yes 9. EateryPlaceMamak=no10. Laundry=yes, 11. MedicalShop=no

(a)

Business Type Attributes Ranking

Mini

market

1. CornerShopLot=yes,2. RoadEntry=yes, 3. Clinic=yes,4. Hardware=no 5. 24HrsConvenience=no 6. FashionAndClothing=no, 7. Hotel=no 8. Bank=no, 9. FitnessCentre=no, 10. Pets=no11. CarAccessories=no, 12. CarWorkshop=no

(b)

Business Type Attributes Ranking

Saloon

1. CornerShopLot=no, 2. FacingRoadView=yes, 3. CarWorkshop=no, 4. Hardware=yes,5. Laundry=no, 6. Pets=no7. RoadEntry=no, 8. CarAccessories=yes,9. EateryPlaceMamak=no,10. Learning=no 11. WellnessCentre=no

(c)

Location Analytics for Optimal Business Retail Site Selection 401

Page 11: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

be operated. Based on the outcome from the association, we will run the Decision Treeclassification to predict the Business Type based on the characteristics. As can beobserved from the decision tree, we summarize the outcome in Table 6:

Based on the summarized information, the Business Type output is in conformancewith the earlier association ranking of characteristics, where the top business such asEatery Place and Mini Market appear in all classification output. Other businesses suchas Bank also appear due to these stores are having the same characteristics as our topbusiness type stores. As a comparison, when we use the same characteristics from theoutput of Association as an input for the Classification, we can see that the output resultoccurred is similar with the Business Type which in our case the Eatery Place and MiniMarket (which serve as the input for the Association). The intention of using bothmethod is different. We used Association to get the Location Characteristics whereasthe intention of using Classification is to get the Business Type.

5 Discussion

5.1 Mapping of Results to Research Questions

Below are some of the discussions based on the research findings.

a. What are the types of businesses that can be found at certain location?We want to gauge the overall data and what information can be extracted. Our datais taken from around twenty different Districts in Klang Valley region. From thedata gathered, we found that there exist around 58 Business Type. Table 7 showsthe top ten businesses (based on count) that occur.

Table 6. Summary of decision tree outcome/result based on selected location characteristics

Attributes Business type

All attributes EateryPlaceMamakEateryPlaceBankMiniMarket

CornerShopLot + RoadEntry + Clinic + Office + Pets +Saloon + CarWorkshop + WellnessCentre + EateryPlaceMamak +Laundry + MedicalShop

EateryPlaceMamakEateryPlaceMiniMarket

CornerShopLot + RoadEntry + Clinic + Hardware +24HrsConvenience + FashionAndClothing + Hotel +Bank + FitnessCentre + Pets + CarAccessories + CarWorkshop

EateryPlaceBankMiniMarket

CornerShopLot + FacingRoadView + CarWorkshop +Hardware + Laundry + Pets + RoadEntry + CarAccessories +EateryPlaceMamak + Learning + WellnessCentre

EateryPlaceMiniMarket

402 A. M. B. M. Rohani and F.-F. Chua

Page 12: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

Based on the outcome, these are the type of stores exist in our data set. The coverageof district as mentioned can be considered as a high population area with manybusiness activities that take place. Based on the outcome of this research, we knowthat each location is having different characteristics that attract certain businesses.However, looking at the data gathered, given an area with similar characteristics, wecan foresee and have an idea of the suitable type of businesses that can be opened inthat particular area.b. What are the top location characteristics that can be reported from the data?

We selected top three Businesses Type with the most number of outlets in order tolist down the top characteristics as shown in Table 8.

All three businesses type will generally have these characteristics existed in theirsurroundings with the related type of shops as part of the attributes. Interestingly inthe combination above, it is observed that attribute CornerShopLot showing a “yes”and “no” value. We need to highlight that these two opposite characteristics arecoming from two different businesses, which is Mini Market and Saloon.

Table 8. Summarized characteristics based on three top business type

Business type Attributes ranking

Eatery placeMini marketSaloon

1. CornerShopLot = yes,2. RoadEntry = no,3. Clinic = yes,4. RoadEntry = yes,5. Hardware = no6. 24HrsConvenience = no7. CornerShopLot = no,8. FacingRoadView = yes,9. CarWorkshop = no,10. Pets = no

Table 7. Top ten businesses from the data set

# Business type Frequency

1 Eateryplace 1252 Mini market 453 Saloon 394 Clinic 375 Learning 356 Carworkshop 347 Office 348 Other mini market 329 Eatery place mamak 2910 Wellness centre 25

Location Analytics for Optimal Business Retail Site Selection 403

Page 13: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

c. Given a set of characteristics of a location, what are the most suitable businessesthat can be opened at the particular area?We can use Decision Tree classification in order to predict the suitable business tobe opened given the attributes of the location. For example, in our case for acombination of characteristics with corner shop lot and a road entry plus a laundryand a motorcycle workshop, the suggested business to be opened is related to:

i. a restaurant (eatery place),ii. grocery shop (mini market) andiii. a polyclinic establishment (clinic).We can run other similar scenario by giving a different input of the characteristicsand will yield another set of suitable business type.

6 Conclusion

This paper presented the methodology of producing the result beginning with the datacollection, data preparation and the mining technique to determine the optimal businesslocation selection. We contribute on the data collection presented using Google Maps.As far as the usage of data online, it is a novel approach whereby data being presentedin such a way where manual records is done in order to capture the physical charac-teristics of a location. Although it’s a manual process of opening and browsing themaps, it presented the opportunity over traditional method in many ways. This paperalso suggested the usage of data mining techniques in order to do some prediction onthe type of businesses that suitable to be opened, given some characteristics. Wepresented the proposed solution on each of the processes and presented the related Rcodes in order to produce the result using the data mining technique. It gives indicationto the interested party on the characteristics that may have more value to the business.For example, we can target the business owner in order to decide where to open theirbusinesses. For future work, this research can be extended using more data from otherinput such as: (1) Financial performance of the stores - to compare whether the physicalcharacteristics having any significance relation with the stores turnover or sales revenue(2) Number of customer of the premises – to study on the traffic of patron to the shop inrelation to the top location characteristics chosen in this research (3) The social mediadata for number of check-ins or likes – to study on the possibility of higher check-ins orlikes for the location with high ranks of characteristics.

References

1. Hernandez, T., Bennison, D.: The art and science of retail location decisions. Int. J. RetailDistrib. Manag. 28(8), 357–367 (2000)

2. Zentes, J., Morschett, D., Schramm-Klein, H.: Store location – trading area analysis and siteselection. In: Strategic Retail Management, pp. 203–225 (2011)

3. Waxman, L.: The coffee shop: social and physical factors influencing place attachment.J. Inter. Des. 31(3), 35–53 (2006)

4. Lin, J., Oentaryo, R., Lim, E.P., Vu, C., Vu, A., Kwee, A.: Where is the goldmine? Findingpromising business locations through Facebook data analytics. In: Proceedings of the 27thACM Conference on Hypertext and Social Media, pp. 93–102. ACM, New York (2016)

404 A. M. B. M. Rohani and F.-F. Chua

Page 14: Location Analytics for Optimal Business Retail Site …...The previous work on selecting location factors are mainly using the traditional qualitative approach of interviewing, observation

5. Karamshuk, D., Noulas, A., Scellato, S., Nicosia, V., Mascolo, C.: Geo-spotting: miningonline location-based services for optimal retail store placement. In: Proceedings of the 19thACM SIGKDD International Conference on Knowledge Discovery and Data Mining,pp. 793–801. ACM, New York (2013)

6. Xu, M., Wang, T., Wu, Z., Zhou, J., Li, J., Wu, H.: Store location selection via miningsearch query logs of Baidu maps (2016). arXiv:1606.03662

7. Agrawal, R., Srikant, R.: Fast algorithms for mining association rules in large databases. In:Proceedings of the 20th International Conference on Very Large Data Bases, pp. 487–499.Morgan Kaufmann Publishers, San Francisco (1994)

8. Koperski, K., Han, J.: Discovery of spatial association rules in geographic informationdatabases. In: Egenhofer, Max J., Herring, John R. (eds.) SSD 1995. LNCS, vol. 951,pp. 47–66. Springer, Heidelberg (1995). https://doi.org/10.1007/3-540-60159-7_4

9. Kouris, I.N., Makris, C., Theodoridis, E., Tsakalidis, A.: Association Rules Mining forRetail Organizations. Encyclopedia of Information Science and Technology, 2nd edn.,pp. 262–267. IGI Global, Hershey (2009)

10. Crewe, L.: Geographies of retailing and consumption. Prog. Hum. Geogr. 24(2), 275–290(2000). https://doi.org/10.1191/030913200670386318

11. Hollander, S.C.: Multinational Retailing. Institute for International Business and EconomicDevelopment Studies, Michigan State University (1970)

12. Fernie, J., Moore, C.M., Lawrie, A.: A tale of two cities: an examination of fashion designerretailing within London and New York. J. Prod. Brand Manag. 7(5), 366–378 (1998)

13. Radon, A.: Luxury brand exclusivity strategies – an illustration of a cultural collaboration.J. Bus. Adm. Res. 1(1), 106 (2012)

14. Jensen, P.: Network-based predictions of retail store commercial categories and optimallocations. Phys. Rev. E 74(3), 035101 (2006)

15. Garber, L.: Analytics goes on location with new approaches. Computer 46(4), 14–17 (2013).https://doi.org/10.1109/MC.2013.123

16. ESRI: Using Location Intelligence to Maximize the Value of BI, 16 November 2011. http://www.cio.in/whitepaper/using-location-intelligence-maximize-value-bi-moved. Accessed 5Aug 2017

17. Angelaccio, M., Buttarazzi, B., Basili, A., Liguori, W.: Using geo-business intelligence toimprove quality of life. In: 2012 IEEE First AESS European Conference on SatelliteTelecommunications (ESTEL), pp. 1–6. IEEE (2012)

18. Shekhar, S., Huang, Y.: Discovering spatial co-location patterns: a summary of results. In:Jensen, Christian S., Schneider, M., Seeger, B., Tsotras, Vassilis J. (eds.) SSTD 2001.LNCS, vol. 2121, pp. 236–256. Springer, Heidelberg (2001). https://doi.org/10.1007/3-540-47724-1_13

19. Niti, D.: Retail location analysis: a case study of Burger King & McDonald’s in Portage &Summit Counties, Ohio. Kent State University (2007)

20. The edge [News], 9 June 2014. http://www.equatoassist1.com/CMS_99SMart2/Admin/uploads/news/839bfa4c-049d-4fa9-9dff-a0b5f9cf1d9a/Pages%20from%2020140609_TEM_1018.pdf. Accessed 30 Aug 2017

21. Fayyad, U., Piatetsky-Shapiro, G., Smyth, P.: From data mining to knowledge discovery indatabases. AI Mag. 17(3), 37 (1996). https://doi.org/10.1609/aimag.v17i3.1230

22. Han, J., Kamber, M., Pei, J.: Mining frequent patterns, associations, and correlations: basicconcepts and methods. In: Han, J., Kamber, M., Pei, J. (eds.) Data Mining, 3rd edn. MorganKaufmann, Boston (2012)

23. Gupta, M., Agarwal, N.: Classification techniques analysis. In: Proceedings of NationalConference on Computational Instrumentation, pp. 120–128 (2010)

Location Analytics for Optimal Business Retail Site Selection 405