Comparison of supervised machine learning techniques for customer churn prediction based on analysis of customer behavior
| Date | 13 March 2017 |
| Published date | 13 March 2017 |
| Pages | 65-93 |
| DOI | https://doi.org/10.1108/JSIT-10-2016-0061 |
| Author | Samira Khodabandehlou,Mahmoud Zivari Rahman |
| Subject Matter | Information & knowledge management,Information systems,Information & communications technology |
Comparison of supervised
machine learning techniques for
customer churn prediction based
on analysis of customer behavior
Samira Khodabandehlou
Master of Information Technology Engineering, Tehran, Iran, and
Mahmoud Zivari Rahman
Allameh Tabataba’i University, Tehran, Iran
Abstract
Purpose –This paper aims to provide a predictive framework of customer churn through six stages for
accurateprediction and preventing customer churnin the field of business.
Design/methodology/approach –The six stages are asfollows: first, collection of customer behavioral
data and preparation of the data; second, the formation of derived variables and selection of influential
variables, usinga method of discriminant analysis; third, selection of trainingand testing data and reviewing
their proportion;fourth, the development of predictionmodels using simple, bagging and boosting versionsof
supervised machine learning; fifth, comparison of churn prediction models based on different versions of
machine-learning methods and selected variables; and sixth, providing appropriate strategies based on the
proposedmodel.
Findings –According to the results, five variables, the number of items, reception of returned items, the
discount, the distribution time and the prize beside the recency, frequency and monetary (RFM) variables
(RFMITSDP), were chosen as the best predictor variables. The proposed model with accuracy of 97.92
per cent, in comparison to RFM, had much betterperformance in churn prediction and among the supervised
machine learning methods,artificial neural network (ANN) had the highest accuracy, and decision trees (DT)
was the least accurate one. The resultsshow the substantially superiority of boosting versions in prediction
comparedwith simple and bagging models.
Research limitations/implications –The period of the available data waslimited to two years. The
research data were limited to only one grocery store whereby it may not be applicable to other industries;
therefore,generalizing the results to other business centersshould be used with caution.
Practical implications –Business owners musttry to enforce a clear rule to provide a prize for a certain
number of purchased items. Of course,the prize can be something other than the purchased item. Business
owners must accept the items returned by the customers for any reasons, and the conditions for accepting
returned items and the deadline for accepting the returned items must be clearly communicated to the
customers. Storeowners must consider a discount for a certain amount of purchase from the store. They have
to use an exponential rule to increase the discount when the amount of purchase is increased to encourage
customers formore purchase. The managers of large stores must try to quicklydeliver the ordered items, and
they should use equippedand new transporting vehicles and skilled and friendly workforce for deliveringthe
items. It is recommended that the types of services, the rules for prizes,the discount, the rules for accepting
the returned items and the method of distributingthe items must be prepared and shown in the store for all
the customers to see. The special services and reward rules of the store must be communicated to the
customersusing new media such as social networks. To predict the customer behaviors based on the data, the
future researchersshould use the boosting method because it increasesefficiency and accuracy of prediction.
It is recommended that for predicting the customer behaviors, particularly their churning status, the ANN
method be used. To extract and select the importantand effective variables influencing customer behaviors,
the discriminant analysis methodcan be used which is a very accurate and powerful method for predicting
the classesof the customers.
Supervised
machine
learning
techniques
65
Received3 October 2016
Revised26 February 2017
Accepted17 April 2017
Journalof Systems and
InformationTechnology
Vol.19 No. 1/2, 2017
pp. 65-93
© Emerald Publishing Limited
1328-7265
DOI 10.1108/JSIT-10-2016-0061
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1328-7265.htm
Originality/value –The current study tries to fill this gap by considering five basic and important
variablesbesides RFM in stores, i.e. prize, discount, accepting returns,delay in distribution and the number of
items, so that the business ownerscan understand the role services such as prizes, discount, distributionand
acceptingreturns play in retraining the customers and preventingthem from churning. Another innovation of
the current study is the comparisonof machine-learning methods with their boostingand bagging versions,
especiallyconsidering the fact that previous studies do not consider the baggingmethod. The other reason for
the study is the conflicting results regardingthe superiority of machine-learning methods in a more accurate
prediction of customer behaviors, including churning. For example, some studies introduce ANN (Huang
et al., 2010;Hung and Wang, 2004;Keramatiet al.,2014;Runge et al., 2014), some introduce support vector
machine ( Guo-en and Wei-dong,2008;Vafeiadis et al.,2015;Yu et al.,2011)and some introduce DT (Freund
and Schapire, 1996;Qureshiet al., 2013;Umayaparvathi and Iyakutti, 2012) as the best predictor, confusing
the users of the resultsof these studies regarding the best prediction method. The currentstudy identifies the
best prediction method specificallyin the field of store businesses for researchers and the owners. Moreover,
another innovation of the current study is using discriminant analysis for selecting and filtering variables
which are important and effective in predicting churners and non-churners, which is not used in previous
studies. Therefore,the current study is unique considering the usedvariables, the method of comparing their
accuracyand the method of selecting effective variables.
Keywords RFM model, Bagging algorithm, Boosting algorithm, Churn prediction,
Supervised machine learning techniques
Paper type Research paper
1. Introduction
Nowadays, due to the advances of technology, the change in selling approaches and the
creation of competitive markets, the markets are characterized by supply surplus, causing
the customer to be considered as the real ruler of the market. Therefore, commercial
enterprises must shift from focusing on the products toward focusing on customers and
manage their behaviors to prevent them from leaving and to obtain the highest profits and
revenues for their own organizations (Dick and Basu, 2003;Lai, 2009;Payne and Frow,
2004). Customer relationship management (CRM) is the process of strict management of
information on the customers and the appropriate management of all the customers to
maximize their loyalty (Kincaid, 2003;Ngai et al.,2009;Parvatiyar and Sheth, 2001;
Umayaparvathi and Iyakutti, 2012). Its main goal is to create satisfaction and happiness in
the customers to prevent them fromchurning, as it is the gravest danger threatening all the
organizations because a small change in the level of customer retention will lead to
significant changes in the shares and the profits of the organization (Agarwal et al.,2004;
Kotler and Keller, 2006;Swift,2001;Van den Poel and Larivie’re, 2004).
Maintaining the customers is the most basic and the most important issue for commercial
entities, particularly, shopping centers. On the other hand, attracting new customers has a very
high cost, sometimes five times the cost of retaining the current customers (Ali and Arıtürk,2014;
Hung and Wang, 2004;Marcus, 1998;Murakani and Natori, 2013;Reichheld and Sasser, 1990;
Reichheld, 1993;Shoemaker and Bowen, 1998;Tamaddoni Jahromi et al.,2014;Umayaparvathi
and Iyakutti, 2012). However, many commercial organizations suffer from losing their valuable
customers to the competition. This is known as customer churning (Huang et al., 2012,2010;
Hung and Wang, 2004;Lu et al.,2014;Umayaparvathi and Iyakutti, 2012;Yu et al., 2011).
Customer churning (customer attrition) is a marketing term in which the customer becomes
interested in another organization or product (Chen, 2016;Coussement and Poel, 2009;Glady et al.,
2009;Hung and Wang, 2004;Keramati et al., 2014;Yu et al.,2011),leadingtoareductioninsale
revenues and an increase in the cost of attracting customers (Coussement and Poel, 2009;
Haywood, 1988). Various studies have been carried out for predicting customer churn and its
functions (Ali and Arıtürk, 2014;Chen, 2016;Hung and Wang, 2004;Keramati et al., 2014;Kim
JSIT
19,1/2
66
To continue reading
Request your trial