Customer Retention Strategies using Machine Learning: SVM Algorithm

This report examines customer retention strategies within the context of increasing competition and the importance of customer satisfaction in driving business success. It highlights the role of Customer Relationship Management (CRM) systems in gaining customers, maintaining relationships, and improving retention. The report emphasizes the value of machine learning models, specifically the Support Vector Machine (SVM) algorithm, in enhancing customer retention strategies. It discusses the concept of customer retention, its relation to customer loyalty and satisfaction, and the calculation of retention rates. Furthermore, the report addresses churn prediction, outlining the types of churn (voluntary and involuntary) and the challenges in the telecommunications industry. It details the SVM algorithm, its function in data analysis for churn prediction, and the framework for its implementation. The report concludes by emphasizing the need for customer retention and the advantages of SVM in predicting churn, providing a comprehensive overview of the topic.

Customer retention strategies using machine
learning
Name of the student
Name of the Unniversity
Desingnation
XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

Abstract— In current days the customers are getting more
attracted towards the quality of service (QoS) provided by the
organizations. However, the current era is evidencing higher
competition in providing technologically advanced QoS to the
customers. Nevertheless, efficient customer relationship
management systems can be advantageous for the organization
for gaining more customers, maintaining customer
relationships and improve customer retention by adding more
profit to the organizational business. Furthermore, the
machine learning models such as support vector machine
algorithms can add more value to the customer retention
strategies.
Keywords—customer relationship management, customer
retention, machine learning, support vector algorithm
I. INTRODUCTION
Customers always play vital role in increasing profit and
revenue of every organizations; hence, to gain customer
satisfaction it is important for the organizational managers
to maintain one efficient customer relationship management
system by selecting the target customers and maintaining
effective relationship with them. Moreover the CRM system
will be helpful for the organization in identifying the most
prominent group of customers and their behavior; which
will become beneficial for the organization in understanding
the retention strategies in a better way. Moreover, higher the
customer loyalty, lesser is the customer churn rate; hence
using machine learning algorithm such as support vector
algorithm can add value in preventing the customer churn.
This report will focus on the customer retention of the with
the usage of support vector machine learning in gaining
customer loyalty and increasing retention.
II. CUSTOMER RETENTION
The vital goal of this process is to help the organizations in
gaining immense customer loyalty by nullifying the number
of customer defections [1]. The concept of customer retention
is directly related to the customer satisfaction and loyalty.
Customer retention is a combine outcome of several things
such as, customer satisfaction, customer delight, customer
relationship management and customer switching cost.
Effective customer retention strategies can surely turned to
be advantageous for the organizations as it is helpful in
gaining long lasting and loyal customer relationship [2].
Hence, it is important for the organizations to calculate the
customer retention rate. If,
A= total number of consumers at the end of the period
B= total number of consumers acquired during the period
C= total number of consumer at the starting period
Then the retention rate can calculated as,
A-B/C*100
The organizations can conduct the following process to
enhance the customer retention process
 Listening the customer demand.
 Considering customer feedback and taking
appropriate action to it.
 Resolving customer queries within a small time
period [3].
 Creating consumer follow up plan.
 Identifying the risk customers.
 Building tactics for wining back the lost customers.
 Anticipating the customer churn.
III. CHURN PREDECTION
Examining the customer attrition rate in an organization
implies the process of churn analysis. In the
telecommunication industries, the churn can be identified as
the number customers who had discontinued their
subscription in a certain time period [4]. A typical churn rate
measures the number of customers moving in and out within
a given time period. Moreover, for the telecommunication
industry, the movement if the customers from one company
to another, called churn [5]. The current scenario is
evidencing a higher number of churn customers as the
particular industry is trying hard to retain more profitable
customers. Moreover, the churn can be classified into two
types. In case of nonpayment of the bills, fraud activity or
any such activity, when the industry itself decides to remove
the consumers, it is named as involuntary churn. In contrast
to this, when the customer intended to change or leave the
organization, the particular activity becomes a voluntary
churn [6]. However, In case of the telecom industry, the
continuous increase in the number of service providers,
becoming the vital reason for creating more churn customer
for the companies. Nevertheless, understanding the
customer demand and gaining loyal customer can eliminate
the churn in a higher rate.
IV.SUPPORT VECTOR MACHINE LEARNING
ALGORITHM
Machine learning can be considered as the effective
application of the artificial intelligence, which has been
widely used by the telecom industries in evaluating and
nullifying the customer churn. Support vector machine
learning is one vital machine learning algorithm that
efficiently performs the data analysis for predicting the
churn [7]. Moreover, the support vector machine (SVM)
algorithm encompasses with a series of supervised learning
methods for separating the data points [8]. The support
vector machine works by mapping the data in order to create

hyper planes. The optimal hyper plane in the support vector
machine can be described as follows,
(Figure 1 [9])
Furthermore, those data points are mapped to higher
dimensional space for identifying the suitable hyper planes
among the instances of several classes. The new instances
are classified into specific classes depending upon their
proximity. The mapping functions used in this algorithm are
generally obtain from a combination of labeled training data
set, such as, {xi, yi}i = 1,…….,n. hence, the modeling
objective for finding the linear decision function can be
described as follows;
(1)
In the above equation,
w = weight vector
ϕ= nonlinear mapping function
b= constant
This particular regression problem is formulated for
minimizing the below risk function
In the above equation,
ε= intensive loss function and,
(2)
After introducing the slack variables,
Minimize:
(3)
Subject to:
(4)
In the above, Cis the regularized constant having value
greater than 0. This constant is taken to make a balance
between the model flatness and training error. Moreover,
this constant is responsible for representing penalty for any
kind of predection error. The objective of support vector is
to minimize w2. Hence, the above equations can be changed
by means of the lagrangian multiplies into one quadratic
equation as follows;
(5)
K(.) represents a Kernel function, which value calculated by
the inner products of the vectors xi and xj , which
represented by the future space ϕ(xi ) and ϕ(xj)
α i and αi* represents the lagrange multiplies. These
multipliers are subjected to the following;
Furthermore, the kernel function satisfies Mercer’s
condition. Hence,
There exist several useful kernel functions; however, the
radial basic kernel function will become beneficial in
classifying the customer churn in the telecommunication
industries.
This radial basic kernel function is represented by
The correctness of a SVM model depends on the efficient
selection of the associated model parameters. A list of
parameters such as error trade off parameter (C), parameter
for controlling the width of intensive zone and many others,
which needs cross validation.
Support vector machine algorithm is one of the
powerful prediction method for identifying the churn rate. In
contrast to the traditional churn prediction methods, SVM

allows the problem solution to depend upon the subsets of
the data set, which rovides comparative computational
advantages to the technique [10]. Apart from that, instead of
minimization of training error, the support vectom machine
algorithm focuses on minimizing the generalization error.
Adoption of this technique is turning out to be the vital
impeding factor for the churn prediction in the telecom
industries. The following image shows the SVM based
framework in order to predict the churn.
(Figure 2: churn prediction framework)[9]
V. CONCLUSION
Therefore, from the above discussion, it can be concluded
that, regardless of the type of organization, every
organization need to concern about the customer churn.
Customer retention is the process of maintaining the loyalty
of the customer by understanding the customer demand and
serving them accordingly. Powerful churn prediction model
will help the organizational management to predict the
customer churn. Depending on the complex data of the
telecommunication industry, support vector machine can
turned out advantageous for predicting the churn rate. The
above report has focused on the concept of customer
retention along with the churn prediction. Apart from that,
the use of support vector machine in order to enhance the
churn prediction process has discussed here along with the
algorithm.
VI.REFERENCES
[1] Siu NY, Zhang TJ, Yau CY. The roles of justice and
customer satisfaction in customer retention: A lesson from
service recovery. Journal of business ethics. 2013 Jun
1;114(4):675-86.
[2] Hossain MM, Suchy NJ. Influence of customer
satisfaction on loyalty: A study on mobile
telecommunication industry. Journal of Social Sciences.
2013;9(2):73-80.
[3] Maldonado S, Flores Á, Verbraken T, Baesens B, Weber
R. Profit-based feature selection using support vector
machines–General framework and an application for
customer retention. Applied Soft Computing. 2015 Oct
1;35:740-8.
[4] Maga M, Canale P, Bohe A, inventors; Accenture Global
Services Ltd, assignee. Churn prediction and management
system. United States patent US 8,712,828. 2014 Apr 29.
[5] Vafeiadis T, Diamantaras KI, Sarigiannidis G,
Chatzisavvas KC. A comparison of machine learning
techniques for customer churn prediction. Simulation
Modelling Practice and Theory. 2015 Jun 1;55:1-9.
[6] Haenlein M. Social interactions in customer churn
decisions: The impact of relationship directionality.
International Journal of Research in Marketing. 2013 Sep
1;30(3):236-48.
[7] Farquad MA, Ravi V, Raju SB. Churn prediction using
comprehensible support vector machine: An analytical CRM
application. Applied Soft Computing. 2014 Jun 1;19:31-40.
[8] Vafeiadis T, Diamantaras KI, Sarigiannidis G,
Chatzisavvas KC. A comparison of machine learning
techniques for customer churn prediction. Simulation
Modelling Practice and Theory. 2015 Jun 1;55:1-9.
[9] Rodan A, Faris H, Alsakran J, Al-Kadi O. A support
vector machine approach for churn prediction in telecom
industry. International journal on information. 2014 Aug
1;17(8):3961-70.
[10] Brandusoiu I, Toderean G. Churn prediction in the
telecommunications sector using support vector machines.
Margin. 2013;1:x1.