Customer Churn Prediction using Association Rule Mining

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 3 Issue 5, August 2019 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
@ IJTSRD | Unique Paper ID – IJTSRD26818 | Volume – 3 | Issue – 5 | July - August 2019 Page 1886
Customer Churn Prediction using Association Rule Mining
Mie Mie Aung, Thae Thae Han, Su Mon Ko
Information Technology Support and Maintenance Department,
University of Computer Studies, Meiktila, Myanmar
How to cite this paper: Mie Mie Aung |
Thae Thae Han | Su Mon Ko "Customer
Churn Prediction using Association Rule
Mining" Published
in International
Journal of Trend in
Scientific Research
and Development
(ijtsrd), ISSN: 2456-
6470, Volume-3 |
Issue-5, August
2019, pp.1886-1890,
https://doi.org/10.31142/ijtsrd26818
Copyright © 2019 by author(s) and
International Journal ofTrend inScientific
Research and Development Journal. This
is an Open Access article distributed
under the terms of
the Creative
Commons Attribution
License (CC BY 4.0)
(http://creativecommons.org/licenses/by
/4.0)
ABSTRACT
Customer churn is one of the most importantmetrics fora growing businessto
evaluate. It is a business term used to describe the loss of clients orcustomers.
In the retail sales and marketing company, customers havemultiplechoices of
services and they frequently switch from one service to another. In these
competitive markets, customers demand best products and services at low
prices, while service providers constantly focus on getting hold of as their
business goals. An increase in customer retention of just 5% cancreateatleast
a 25% increase in profit. Therefore, customer churn rate is importantbecause
it costs more to acquire new customers than it does to retain existing
customers. In this paper, we apply the method to the retail sales and
marketing company customer churndata set.This paper provides anextended
overview of the literature on the use of data mining in customer churn
prediction modeling. It will help the retail sales and marketing company to
present the targeted customers with the estimatedloss ofclients orcustomers
for the promotion in direct marketing.
KEYWORDS: Data Mining, Customer Churn Prediction, Association Rule Mining,
FP-Growth
1. INTRODUCTION
Data mining (DM) methodology has a tremendous
contribution for researchers to extract the hidden
knowledge and information whichhavebeeninheritedin the
data used by researchers and it is to extract the knowledge
and information which have been hidden in a large volume
of data. The rapid growth of the market in every sector is
leading to a bigger subscriber base for service providers.
Service providers have realized the importance of the
retention of existing customers. Satisfying customer’s needs
is the key for business success. Customer Relationship
Management (CRM) is a business strategy that aims to
understand, anticipate and manage the needs of an
organization’s current and potential customers. Customer
retention has become a significant stage in CRM, which is
also the most important growth point of profit. Retail Sales
and Marketing across the world are approaching saturation
levels. Therefore, the current focus is tomovefromcustomer
acquisition towards customer retention.
In this paper, we apply the FP-Growth method to the retail
sales and marketing company customer churn data set. One
of the currently fastest and most popular algorithms for
frequent item set mining is the FP-growth algorithm. It is
based on a prefix tree representation of the given database
of transactions (called an FP-tree), which can save
considerable amounts of memory for storing the
transactions. Data mining techniques are used toimplement
customer classification in CRM because mass volumeofdata
is needed to analyze by implementing an efficient and
effective Association Rule Mining based technique. FP-
Growth is used to find the number of customers churns.
Customer churn is the action of the customer who is like to
leave the company and it is one of the mounting issues of
today’s rapidly growing and competitive the retail sales and
marketing company. To minimize the customer churn,
prediction activity to be an important part of the retail sales
and marketingcompany’s vitaldecisionmakingandstrategic
planning process.
1.1 Churn Prediction
Today numerous the retail sales and marketing companies
areprompt all over theworld. The retail sales and marketing
company is (facing a severe)lossofrevenueduetoincreasing
competition among them and loss of potential customers.
Churn is the activity of the retail sales and marketing
company is the customers leaving the current company and
moving to another company. Manycompaniesarefindingthe
reasonsoflosingcustomersbymeasuringcustomerloyaltyto
regain the lost customers. To keep up with the competition
and to acquire as many customers, most operators invest a
huge amount of revenue to expand their business in the
beginning. In the retail sales and marketing company each
company provides the customers with huge incentives to
attract them to change to their services, it is one of the
reasons that customer churnis a big probleminthecompany
nowadays. To prevent this, the company should know the
reasons for which the customer decides to move on to
another company. The Churns canbeclassifiedintotwomain
categories: InvoluntaryandVoluntary.Involuntaryareeasier
to identify. Involuntary churn is those customers whom the
IJTSRD26818

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
retail sales and marketing company decides to remove as a
subscriber. They are churned for fraud, non-payment and
those whodon‘t use theservice.Voluntarychurnisdifficultto
determine because it is the decision of the customer to
unsubscribe from the service provider. Voluntary churn can
further be classified as incidental and deliberate churn . The
former occurs without any prior planning by the churn but
due to change in the financial condition, location, etc. Most
operators are trying to deal with these types of churns
mainly.
1.2 Churn Management
Churn management is very important forreducing churns as
acquiring a new customer is more expensive than retaining
the existing ones. Churn rate is the measurement for the
number of customers moving out and in during a specific
period of time. If the reason for churning is known, the
providers can then improve their services to fulfill the needs
of the customers. Churns can be reduced by analyzing the
past history of thepotentialcustomerssystematically.Alarge
amount of information is maintained by the retail sales and
marketing company for each of their customers that keep on
changing rapidly due to a competitive environment. The
information includes the details about billing, calls and
network data. The huge availability of information arises the
scope of using Data mining techniques in the company’s
database. The information available can be analyzed in
different perspectives to provide various ways to the
operators to predict and reduce churning. Only the relevant
details are used in the analysiswhichcontributestothestudy
from theinformation given. Data mining techniquesareused
for discovering the interesting patterns within data and it
helps to learn to predict whetheracustomerwillchurnornot
based on customer‘s data stored in the database.
2. RELATED WORKS
Berry and Linoff (2000) defines data mining as the process
of exploring and analyzing huge datasets, in order to find
patterns and rules which can be important to solve a
problem. Berson et al. (1999); Lejeune extract or detect
hidden patterns or information from large databases. Data
mining is motivated by the need for techniques to support
thedecision maker in analyzing, understanding and
visualizing the huge amounts of data that have been
gathered from business and are stored in data warehouses
or other information repositories. Data mining is an
interdisciplinary domain that gets together artificial
intelligence, database management, machine learning, data
visualization, mathematic algorithms, and statistics data
mining is considered by some authors as the core stage of
the Knowledge Discovery in Database (KDD) process and
consequently it has received by far the most attention in the
literature (Fayyad et al., 1996a). Data mining applications
have emerged from a variety of fields including marketing,
banking, finance, manufacturing and health care (Brachman
et al., 1996). Moreover, data mining has also been applied to
other fields, such as spatial, telecommunications, web and
multimedia.
3. THEORETICAL BACKGROUND
Data Mining is very famous technique for churn prediction
and it is used in many fields. It refers to the process of
analyzing data in order to determine patterns and their
relationships. It is an advanced technique which goes deep
into data and uses machine learning algorithms to
automatically shift through each record and variable to
uncover the patterns and information that may have been
hidden. Data mining is used to solve the customer churn
problem by identifying the customer behavior from large
number of customer data. Its techniques have been used
widely in churn prediction context such as Support Vector
Machines (SVM), Decision Tree (DT), Artificial Neural
Network (ANN) and Logistic regression.
3.1 Customer Churn Prediction Model
Customer Relationship Management (CRM) system have
been developed and it is applied in order to improve
customer acquisition and retention. Increase of profitability
and to support important analytical tasks such as predictive
modeling and classification; CRM applications hold a huge
set of information regarding each individual customer. This
information is gained from customers’ activity at the
company, data entered by the customer in the process of
registration. The size of gathered data is usually very large,
which results in high dimensionality, making to analyze a
complex and challengingtask.Therefore, beforebeginningto
use a churn prediction method a data reduction techniqueis
used, deciding with application domain knowledge which
attributes can be of use and which can be ignored. Missing
values should also be regarded – on attribute level these can
be ignored if they are with low significance, whereas on
record level they have to be replaced with a reasonable
estimate. Providing a good estimate for these missingvalues
is an important issue for proper churn prediction.
Figure.1 Customer Churn Prediction Model
3.2 Association Rule Mining
Association rule mining, one of the most important and well
researched techniques of data mining, was first introduced
in. It aims to extract interesting correlations, frequent
patterns, associations or casual structures among sets of
items in the transaction databases orotherdata repositories.
Association rules are widely used in various areas such as
telecommunication networks,marketandrisk management,
inventory control etc. Various associationminingtechniques
and algorithms will be briefly introduced and compared
later. Association rule mining is to find out association rules
that satisfy the predefinedminimumsupportandconfidence
from a given database. The problem is usually decomposed
into two subproblems. One is to find those itemsets whose
occurrences exceed a predefined threshold in the database;
those itemsets are called frequent or large itemsets. The
second problem is to generate association rules from those
large itemsets with the constraints of minimal confidence.
Suppose one of the large itemsets is Lk, Lk = {I1, I2, … , Ik},

association rules with this itemsets are generated in the
following way: the first rule is {I1, I2, … , Ik-1}⇒ {Ik}, by
checking the confidence this rule can be determined as
interesting or not. Then other rule are generated by deleting
the last items in the antecedent and inserting it to the
consequent, further the confidences of the new rules are
checked to determine the interestingness of them. Those
processes iterated until the antecedent becomes empty.
Since the second subproblem is quite straight forward,most
of the researches focus on the first subproblem. The first
sub-problem can be further divided into two sub-problems:
candidate large itemsets generation process and frequent
itemsets generation process. We call those itemsets whose
support exceed the support threshold as large or frequent
itemsets, those itemsets that are expected or have the hope
to be large or frequent are called candidate itemsets.
Association Rule Mining can be viewed as a two-step
process:
1. Find all frequent item sets
 Apriori Method
 FP Growth Method (Frequent Pattern)
2. Generate strong association rules from thefrequent
item sets:
 By definition, these rules must satisfyminimumsupport
and minimum confidence
3.3 Basic Concepts & Basic Association Rules
Algorithms
Let I=I1, I2, … , Im be a set of m distinct attributes, T be
transaction that contains a set of items such that T ⊆ I,Dbea
database with different transaction records Ts. An
association rule is an implication in the form of X⇒Y, where
X, Y ⊂ I are sets of items called itemsets, and X ∩ Y =∅. X is
called antecedent while Y is called consequent, the rule
means X implies Y. There are two important basic measures
for association rules, support(s) and confidence(c).Sincethe
database is large and users concern about only those
frequently purchased items, usually thresholds of support
and confidence are predefined by users to drop those rules
that are not so interesting or useful. The two thresholds are
called minimal support andminimalconfidencerespectively.
Support(s) of an association rule is defined as the
percentage/ fraction of records that contain X ∪ Y to the
total number of records inthedatabase.Supposethesupport
of an item is 0.1%, it means only 0.1 percent of the
transaction contain purchasing of this item.Confidenceof an
association rule is defined as the percentage/fraction of the
number of transactions that contain X ∪ Y to the total
number of records that contain X. Confidenceis ameasure of
strength of the association rules, suppose the confidence of
the association rule X⇒Y is 80%, it means that 80% of the
transactions that contain X also contain Y together. In
general, a set of items (such as the antecedent or the
consequent of a rule) is called an itemset. The number of
items in an itemset is called the length of an itemset.
Itemsets of some length k are referred to as k-itemsets.
Generally, an associationrules mining algorithmcontains the
following steps:
 The set of candidate k-itemsets is generated by 1-
extensions of the large (k -1)-itemsets generated in the
previous iteration.
 Supports for the candidatek-itemsetsaregenerated bya
pass over the database.
 Itemsets that do not have the minimum support are
discarded and the remaining itemsets are calledlargek-
itemsets.
This process is repeated until no more large itemsets are
found. The AIS algorithm was the first algorithm proposed
for mining association rule. In this algorithm only one item
consequent association rules are generated, which means
that the consequent of those rules only contain one item, for
example we only generate rules like X ∩ Y⇒Z but not those
rules as X⇒Y∩ Z. The main drawback of the AIS algorithm is
too many candidate itemsets that finally turned out to be
small are generated, which requires more space and wastes
much effort that turned out to be useless. At the same time
this algorithm requires too many passes over the whole
database.
Apriori is more efficient during the candidate generation
process. Apriori uses pruningtechniques toavoidmeasuring
certain itemsets, while guaranteeing completeness. These
are the itemsets that the algorithm can prove will not turn
out to be large. However there are two bottlenecks of the
Apriori algorithm. One is the complex candidate generation
process that uses most of the time, space and memory.
Another bottleneck is the multiple scan of the database.
Based on Apriori algorithm, many new algorithms were
designed with some modifications or improvements.
3.4 Frequent Pattern Growth (FP Growth)
Finding frequent item sets without candidate generation
1. First, compress the database representing frequent
items into a frequent pattern tree or Data classification
is a two-step process. In the first FP tree, which retains
the itemset association information. FP-tree is an
extended prefix-tree structure storing crucial,
quantitative information about frequent patterns. Only
frequent length-1 items will have nodes in the tree, and
the tree nodes are arranged in such a way that more
frequently occurring nodes will have better chances of
sharing nodes than less frequently occurring ones. FP-
Tree scales much better than Apriori because as the
support threshold goes down, the number as well as the
length of frequent itemsets increase dramatically. The
candidate sets that Apriori must handle become
extremely large, and the pattern matching with a lot of
candidates by searching through the transactions
becomes very expensive. The frequent patterns
generation process includes two sub processes:
constructing the FT-Tree, and generating frequent
patterns from the FP-Tree. Theminingresultisthesame
with Apriori series algorithms. To sum up, the efficiency
of FP-Tree algorithm account for threereasons.Firstthe
FP-Tree is a compressed representation of the original
database because only those frequent items are used to
construct the tree, other irrelevant information are
pruned. Secondly this algorithm onlyscans thedatabase
twice. Thirdly, FP-Tree uses a divide and conquer
method that considerably reduced the size of the
subsequent conditional FP-Tree.
2. Then devide the compressed database into a set of
conditional databases ( a special kind of projected
database), each associated with one frequent item or
“pattern fragment”, mines each such database
separately.

No Variable Name Description
1 Age, Gender, Occupation Demographic variables considered
2 The number of purchase Identifies the number of customer is purchased
3 Frequently used purchase Identifies the most frequently purchase by the consumer
4 Churn Identifies whether customer have changed company or not
5 Product innovation Determines whether product innovation is necessary for sustaining customers
6 Product purchase amount (DpM) Approximates the amount used to purchase product a month
7 Credit purchase amount (CpM) Approximates the amount used to purchase call credits a month
8 Tariffs The type of customer, whether a prepaid or post-paid customer
9 Tenure Length of time a customer has been with a particular subscriber
Table1: The Variables Used In Dataset for This Research
3.5 FP-growth Algorithm
In this section we examine the FP-growth algorithm over a
hypothetical dataset for a sailing company. This example is
picked up from the textbook Data-Mining Concepts and
Techniques (Han & Kamber., 2006). The dataset is a
collection of transaction records. Each transaction has a
unique ID and each item is represented by an index Ij. The
dataset is represented in Table 1. The algorithm starts with
the first scan of the database which derives the set of
frequent items (1-itemsets) and their support counts
(frequencies). Let the minimumsupport countis 2.Thesetof
frequent items is sorted in the order of descending support
count. This resulting set or list is denoted as L. Thus, we
have:
L = {I2: 7, I1: 6, I3: 6, I4: 2, I5: 2}
TID List of items Ids
T100 I1, I2, I5
T200 I2, I4
T300 I2, I3
T400 I1, I2, I4
T500 I1, I3
T600 I2, I3
T700 I1, I3
T800 I1, I2, I3, I5
T900 I1, I2, I3
Table2: Transactional Data for a Sailing Company
An FP-tree is then constructed as follows. First, create the
root of the tree, labeled with “null”. ScandatabaseDa second
time. The items in each transaction are processed in L order
(i.e., sorted according to descending support count), and a
branch is created for each transaction.
Figure2: An FP-tree registers compressed, frequent
pattern information.
The tree obtained after scanning all of the transactions is
shown in Figure 1 with the associated node-links. In this
way, the problem of mining frequentpatterns indatabases is
transformed to that of mining the FP-tree. The FP-tree is
mined as follows: Start from each frequent length-1 pattern
(as an initial suffix pattern); constructitsconditional pattern
base (a “sub database” which consists of the set of prefix
paths in the FP-tree co-occurring with the suffix pattern),
then construct its (conditional) FP-tree, and perform mining
recursively on such a tree. Mining of the FP-tree is
summarized in Table 3.
Item
Conditional
Pattern Base
Conditional
FP-tree
Frequent
Pattern
15
{{I2,I1:1},
{I2,I1,I3:1}}
<I2:2,I1:2> {I2,I5:2},
{I1,I5:2},
{I2,I1,I5:2}
14
{{I2,I1:1},
{I2:1}}
<I2:2> {I2,I1:2}
13
{{I2,I1:2},
{I2:2},
{I1:2}}
<I2:4,I1:2>,
<I1:2>
{I2,I3:4},{I
1,I3:4},{I2,I
1,I3:2}
12 {{I2:4}} <I2:4> {I2,I1:4}
Table3: Mining the FP-tree by creating conditional
(sub-) pattern bases
4. CONCLUSION
This paper deals with the customer churn analysis and
predicting the most profitable customer in the retail sales
and marketing system. Customer churn is one of the most
important metrics for a growing business to evaluate. As
churn management is a major task for companies to retain
valuable customers, the ability to predict customer churn is
necessary. This paper mainly focused on the customer
classification and prediction in Customer Relationship
Management concerned with data mining based on FP
Growth technique. This technique is usedtofindingfrequent
item sets without candidate generation.
References
[1] A Lemmens, & S. Gupta, “Managing Churn to Maximize
Profit”, Harvard Business Schol Working Paper, (14-
020), (2013).
[2] A. Fazlzadeh, M. M. Tabrizi, & K. Mahboobi, “Customer
Relationship Management in Small-Medium
Enterprises”, the case of science and technology parks
of Iran, African Journal of Business
Management.5(15),6160-6168,(2011).

[3] C. Rygielski, J. C. Wang, & D. C. Yen, “Data Mining
Technique for Customer Relationship Management”,
Technology in society, 24(4), 483-502(2002).
[4] D. Pyle, “Data Preparation for Data Mining”, Morgan
Kaufmann Publishers, Los Altos, California, (1999).
[5] Sharma, D. Panigrahi, & P. Kumar, “A Neural Network
Based Approach for Predicting Customer Churn in
Cellular Network Services”, arXiv preprint arXiv:
1309.3945,(2013).
[6] Jiawei Han, Jian Pei, Yiwen Yin: Mining Frequent
Patterns without Candidate Generation in Proceedings
of the 2000 ACM SIGMOD international Conference on
Management of Data (Dallas, Texas, United States,May
15-18, 2000). SIGMOD’00. ACM Press,New York,NY,1-
12.
[7] V. Umayaparvathi and K. Iyakutti, "A Survey on
Customer Churn Prediction in Telecom Industry:
Datasets,Methods andMetrics,"International Research
Journal of Engineering and Technology(IRJET),vol.03,
no. 04, April 2016.

Customer Churn Prediction using Association Rule Mining

More Related Content

What's hot (19)

Similar to Customer Churn Prediction using Association Rule Mining (20)

More from ijtsrd (20)

Recently uploaded (20)

Customer Churn Prediction using Association Rule Mining