Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68

RESEARCH PAPER

Improving the Omnichannel Customers’ Lifetime Value Using

Association Rules Data Mining: A Case Study of Agriculture Bank
of Iran
Mohammad Rezaei1, Ali Sanayei 2, Seyed Fathollah Amiri Aghdaie3, Azarnoush Ansari4

1. PhD Candidate of Business Management, Faculty of Administrative Sciences and Economics,

University of Isfahan, Isfahan, Iran
2. Professor, Department of Management, Faculty of Administrative Sciences and Economics,
University of Isfahan, Isfahan, Iran
3. Associate Professor, Department of Management, Faculty of Administrative Sciences and
Economics, University of Isfahan, Isfahan, Iran
4. Assistant Professor, Department of Management, Faculty of Administrative Sciences and
Economics, University of Isfahan, Isfahan, Iran

(Received: November 30, 2020; Revised: April 23, 2021; Accepted: May 2, 2021)

Abstract
Multi-channel marketing causes the customer to lack a unique identity in different channels. This issue
overshadows the synergy of the channels in strengthening the positive attitude of the customers.
However, an omnichannel marketing strategy can work properly. The main purpose of this study,
which was conducted in Agriculture Bank of Iran, was to develop a comprehensive model for
calculating customers’ lifetime values, analyzing customers’ behaviors in different channels by
association rules data mining, and analyzing the relationship between omnichannel strategy and CLV.
First, the association rules in the big data of customers’ banking transactions in different channels
were identified using association rules data mining. Then, the CLV indicators were identified and
prioritized using interviews, questionnaires, and AHP methods, and the lifetime values of omnichannel
and other customers were calculated and compared using t-test. Then, omnichannel customers were
categorized based on the association rules and the lifetime values of omnichannel customers of
different categories was compared using ANOVA method. Eleven association rules regarding the use
of banking channels by omnichannel customers were identified. The results show that there is a
significant difference between the lifetime values of omnichannel customers and other customers and
the lifetime values of omnichannel customers is 134% more.

Keyword: Omnichannel marketing, Omnichannel banking, CLV, Association rule data mining, Big data.

1. Introduction

In today’s competitive markets, where the products and services life cycle and the duration of
comparative advantages stability are becoming shorter, focusing on valuable customers and
creating a long lasting and valuable communication with them, calculating and improving the
customers’ lifetime value (CLV) have become increasingly important.
CLV, as the present value of all future profits obtained from a customer over the life of his
or her relationship with a firm (Gupta et al., 2006), is very important in developing a
marketing strategy, and strategies based on it can lead to increased profitability for a firm
(Valenzuela et al., 2014).


Corresponding Author, Email: [email protected]
50 Rezaei et al.

On the other hand, with the advent of new technologies, tremendous changes have taken
place in customer service channels. The role of physical retail stores is changing; there is a
slight difference between different offline and online channels such as physical stores and
exhibitions, web stores, mobile apps, and social media, and the development of new channels
crosses old boundaries such as geographical boundaries and customer ignorance (Saghiri et
al., 2017). Customer access to various internet and mobile channels and offline or online
touchpoints and the widespread acceptance of new technologies have reduced barriers to
communication between customers and organizations. These changes are also evident in the
banking market, where banks use multiple and different channels such as ATMs, telephone
banks, internet banks, mobile bank, etc., to provide services to customers. Diversity, breadth,
inconsistency, and lack of integration in different and multiple channels confuse customers in
receiving a unique message from communication channels or a product with a unique feature
and quality from distribution channels and make it difficult for marketers to predict and guide
customer behavior and enhance their lifetime value (Verhoef et al., 2015).
In the meantime, omnichannel marketing strategy has been proposed as a solution to solve
these problems. This strategy creates integration between different and multiple communication
and distribution channels in order to meet the needs and interests of customers by creating
synergy between channels.
On the other hand, the ultimate goal of designing and implementing any business strategy,
including omnichannel strategy, depends on how effective it is in attracting, retaining, and
meeting customer needs and interests, as well as improving customer lifetime value (CLV). A
strategy that does not increase the customer lifetime value cannot lead to the achievement of
the company’s mission and goals. This issue and the specific behavior of omnichannel
customers in the use of channels have received less attention in previous studies.
Applied studies on omnichannel retail are constantly increasing. Shen et al. (2018) found that
76% of market leaders considered omnichannel strategy as a key priority in business, and that
omnichannel management was ranked as the third most important topic in research searches.
Despite various studies on omnichannel marketing, few studies have been done on
monitoring and aggregating omnichannel customers’ behaviors in multiple channels and thus
calculating their lifetime value. Providing a comprehensive model to identify and calculate
customers’ lifetime value, analyzing customer behavior in different channels by association
rules data mining (ARDM), and analyzing the relationship between omnichannel strategy and
customers’ lifetime value – which has been neglected in previous studies – are the innovations
of the present study.
This study, which was conducted as a case study in Agriculture Bank of Iran, seeks to
answer the following questions.
1) Does the implementation of omnichannel marketing strategy improve the lifetime value
of customers?
2) What is the optimal behavior of omnichannel customers based on the sequence of their
use of different banking channels?
So, in this study, first the association rules in customers’ behaviors have been identified
using data mining. Then a comprehensive model has been developed to calculate the lifetime
value of bank customers, the lifetime value of these customers have been calculated and
compared using this model, and the optimal behavior of omnichannel customers, in terms of
the sequence of using different channels, has been identified.
Findings of this study can be used by researchers to identify the various dimensions and
aspects of omnichannel marketing, develop models for calculating CLV, and identify the
relationships between these variables and other variables of marketing science.
In addition, business activists – especially bankers – can use these findings to develop and
Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68 51

implement the effective omnichannel marketing strategy and emphasize the channels whose
sequence of use helps improve customers’ lifetime value in their shopping journey.

2. Literature Review

2.1. Omnichannel

The world of retailing has changed dramatically in the last decade. The advent of online
channels and new digital channels such as mobile and social media channels has changed
retail business models, retail mix, and buyers’ behaviors. With the expansion of various
channels, customers buy not only from physical stores or online, but also through various
channels. For example, they search and retrieve information from one channel and complete
the purchase through another channel (Lee et al., 2019).
Because of these changes in customer behavior, marketing strategies have also changed. In
this regard, marketing strategies are divided into single-channel, multichannel, cross-channel,
and omnichannel strategies in terms of using limited or multiple channels to communicate or
present the product to customers.
Single Channel: The most basic method of communication between businesses and
customers is single channel. In this method, customers communicate with the business
through only one channel, which is mainly physical communication (e.g., store, bank branch,
insurance agency, etc.).
Multichannel: Multichannel marketing is a marketing strategy that uses a combination of
consumer interactive channels such as websites, mail order catalogs, direct mail, email,
mobile, etc.
In the case of banks, the emergence of channels such as ATMs, telephone banks, Internet
banks, mobile banks, etc., has led to the communication of customers with banks in a
multichannel manner.
Multichannel studies show that in many cases single channels of multichannel systems still
try to operate independently to optimize their output, while consumers prefer to choose their
preferred channels based on indicators such as technological factors. As a result, multichannel
systems do not create synergy of parallel supply channels (Bhatnagar & Syam, 2014).
In this strategy, with different independent channels, disproportion and instability of
production and ordering of information as well as poor inventory efficiency are very likely
(Saghiri et al., 2017).
Cross-Channels: This method is more complete than multichannels. In the cross-channel
method, each customer has a unique identity that is identified by the same identity in all
channels. It should be noted that in the cross-channel method, the communication channels
are still independent of each other.
Omnichannel: With the rapid development of in-store technology, multichannel and cross
channel services are changing to omnichannel ones. Omnichannel retail refers to a kind of
retailing in which, regardless of the channel or stage in which customers are located during
the purchase process, the synergetic integration is used to create an integrate brand experience
for customers (Cummins et al., 2016). With the integration of different parallel channels,
omnichannel services provide customers with an integrated, seamless, and cross-channel
shopping experience (Shen et al., 2018).
In such a system, consumers, in their shopping experience, can easily move from one
channel to another and thus find their desired product in one channel (for example, on the
manufacturer’s website), order through another channel (for example, online retailers), and
52 Rezaei et al.

receive the product through another channel (for example, home delivery) (Chopra, 2016;
Verhoef et al., 2015).
The findings of previous studies suggest various benefits for retailers and customers,
including increased cross-selling channels (Cao & Li, 2015; Gallino & Moreno, 2014),
improved operational efficiency and improved customer experience (Gallino & Moreno,
2014), increased customer’s loyalty (Van Baal, 2014) and their trust in retailers (Gallino &
Moreno, 2014). While increasing the integrity of different channels, it increases operational
complexity (Gallino & Moreno, 2014).
Furthermore, by accepting the omnichannel business model, retailers can use a wide range
of technologies to track customer behavior, in both physical and virtual environment, gain
more comprehensive knowledge about each customer, and better shape their shopping
experience (Chen et al., 2018).
In addition, omnichannel helps make the necessary arrangements in the separate service
processes and technologies of different channels in order to create a permanent and integrated
experience for the customers. The integration quality of the channel and the perceived fluency
are the main features of the omnichannel business. The quality of channel integration is an
object-based belief because it demonstrates the ability of omnichannel technology to create
coherence and integration across parallel channels and reflects customer beliefs about
omnichannel technology. Perceived fluency is also a behavioral belief because it refers to how
customers evaluate their inter-channel experience that stems from their actual using behavior
and reflects their beliefs about using omnichannel technology (Shen et al., 2018).

2.2. Omnichannel Banking

Banks operate in a challenging world with rapid technological changes, while their customers
have a perceptual view of technology and their expectations are rising. Such an economic
environment requires financial institutions that review their business strategies and activities
and make changes to the services they provide to customers. Today, banks are gradually
reducing the number of their branches and are designing and implementing processes and
systems that are more efficient and effective. Customers’ visits to branches are also fading
over time; this indicates the need for upgraded and customer-centric processes in self-service
channels to increase and improve branch capabilities and customer satisfaction, and ultimately
lead to improved profits for financial institutions.
In line with these developments, to gain competitive advantages, banks need to move
towards omnichannel banking. Omnichannel banking is different from the current
multichannel banking approach in which banks encourage customers to use the cheapest
channel. Through channels, omnichannel banking can create a sustainable experience for
customers so that they can have seamless visibility of the products and financial services they
need. From a customer perspective, the integration advantage of an omnichannel strategy,
which of course also exists in omnichannel banking, is to increase the value offered by
retailers (Gallino & Moreno, 2014; Gao & Su, 2017).
However, the question is whether the value offered by retailers to customers, which has
been increased by adopting an omnichannel strategy, changes the customer lifetime value. In
order to answer this fundamental question, the concepts of customer lifetime value (CLV) and
association rule must first be explained.

2.3. Customer Lifetime Value (CLV)

Customer Lifetime Value (CLV) is defined as the present value of all future benefits that a
company receives from a customer during its lifetime relationship with the company
Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68 53

(Nikkhahan et al., 2011). According to Fishbein and Ajzen’s (1975) model, intentions allow
individuals to apply all relevant factors that may affect their behavior; in fact, intentions
usually predict customer behavior in the best possible way. Intention or conative component
is between attitude (CLV stimuli) and behavior (CLV), which is known in relational
marketing literature as intentional loyalty (Moliner & Tena, 2016).
Calculating and knowing the CLV enables the company to operate the market
segmentation and resource allocation effectively.
There are several models for measuring CLV, including the RFM model. This model is an
applied model for CLV that is used in various industries to calculate the customers’ lifetime
value. In this model, the recency (R), frequency (F), and monetary (M) parameters are used.
Recency: Recency refers to the days after the last purchase, and means that how many
days ago was the last customer purchase.
Frequency: Frequency refers to the total number of exchanges with the customer and
states that how many times the customer has purchased a service or product from the
company.
Monetary: The monetary parameter means the total amount of money paid and shows
how much the customer has paid in total transactions with the company (Tarokh & Esmaeili
Gookeh, 2019).

2.4. Association Rule Mining

Association Rule Data Mining (ARDM) is one of the most common methods of data mining
that can be used to identify attractive knowledge in a large volume of huge given data set.
Association rules reflect the interdependence between one subject and another. If there is a
connection between several topics, it is possible to predict one of them through other topics.
Today, association rules are used in web data mining, recommender systems, intrusion
detection, marketing, e-commerce, case analysis, risk management, and other disciplines (Shi
et al., 2019). For example, one of the applications of the association rules is their use in the
market basket analysis, through which the buying habits of customers are identified by
analyzing the products that they place in their shopping baskets. For example, this method can
be used with analyzing the products purchased by customers and their sequence over a period
of time. Here, it becomes clear that when customers buy product A, how likely it is that they
will add product B to their product portfolio.
During the process of rule mining, rule is defined as the form of A→ B with two
restrictions of A, B ⊂ I and A ∩ B ≠ ∅, where I represents the set of items. A is a set of items
called antecedent or the left hand side (LHS) and B denotes a set of items referred to as the
consequent of the rule or the right hand side (RHS) (Wang et al., 2019).
The degree of the attractiveness of an association rule is indicated by the functions of
support, confidence, and lift.
The support of an association rule shows the percentage of transactions containing the
union of sets A and B, and it is taken to be the probability, expressed as P(A ∪ B). Confidence
is the proportion of the transactions with A that also contain the union of sets A and B. That is
expressed as Eq.1 (Wang et al., 2019)

(1)

An association rule is considered a valid rule when its support is greater than the minimum
support (ms) threshold defined by the user and its reliability is greater than the minimum
confidence (mc) threshold defined by the user.
54 Rezaei et al.

Lift measure is introduced to make up the shortcoming of confidence measure that ignores
the baseline frequency of the consequent (Eq. 2) (Wang et al., 2019).

(2)

If lift is equal to 1, then A and B are independent. If lift is less than 1, the occurrence of A
is negatively correlated with the occurrence of B. A lift value more than 1 indicates that the
antecedent of a rule positively stimulates the consequent of that rule.
Using the results of ARM in banking, you can understand how much you can expect to use
a channel or other channels (such as POS, mobile banking, etc.) when customers use a
channel (such as ATM).

3. Research Background

In recent years, several studies have been conducted on the importance of omnichannel
marketing, its framework and components, as well as its effects and its relationship with other
variables. and the attention of researchers and business activists to this issue is increasing.
According to Shao (2021), although the adoption of an omnichannel strategy is appropriate
for traditional retailers (e.g., brick and mortar), this strategy may not be effective for online
retailers. In addition, omnichannel retailing does not necessarily lead to lower prices and
customer convenience, and ultimately the omnichannel strategy may not be effective for
manufacturers. Of course, Wagner et al. (2018) believe that retailers can enhance consumer
shopping experiences by providing alternative electronic channels contact points that
participate differently in customers’ online shopping journeys. On the other hand, Shen et al.
(2018) have also pointed out the advantages of omnichannels and believe that the quality of
channel integration significantly affects the perceived validity in different channels. They also
believe that the impact of perceived validity in the use of omnichannel services is decreased
by the experience of internal use and is increased by the experience of external use. Kim and
Chun (2018) have done a more complete study on this; they compared the different strategies
of the production channels and concluded that if the customers are heterogeneous based on
their acceptance of online shopping, the multichannel strategy will be appropriate, and if they
are homogeneous, the best strategy is omnichannel.
Many efforts have been made by researchers to design an omnichannel marketing
framework and influencing factors. Saghiri et al. (2017) designed a conceptual framework for
omnichannel systems, characterized by three dimensions, namely channel stage, channel type,
and channel agent. In this framework, integration and visibility have been examined and
discussed as the main sponsors that support the implementation of the omnichannel
framework. Hsia et al. (2020) also point out that omnichannel retailing is still in its infancy,
and believe that infrastructure, infrastructure synergy, and individual motivations affect the
customer’s positional participation in the positive omnichannel retail experience.
Regarding the factors affecting omnichannel and its effects, Hure et al. (2017) believe that
there is a significant relationship between offline purchase value and online purchase value
with omnichannel purchase value, while no significant relationship is found between mobile
purchase value and omnichannel purchase value. Meanwhile, omnichannel power, which is
itself a function of the integration and perceptual stability variables, acts as an intermediary
variable between the offline purchase value and the omnichannel purchase value. Also Lee et
al. (2019) believe that the dimensions of channel quality integration have a positive effect on
customer commitment, which leads to the positive word of mouth and positive intentions to
buy. With regard to the effects of all channels Herhausen, et al. (2015) believe that channel
Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68 55

integration (resulting from a omnichannel strategy) leads to an increase in the perceptual

quality of services and a reduction in the perceived risk of online shopping.
Various studies have been conducted on customer lifetime value. The International Data
Corporation (IDC) has found that customers who use both online and physical channels have
30 percent more lifetime value than customers who use single channels (online or physical)
(Lee et al., 2019). In addition, Suárez and Tejero (2021) have shown that banks make
decisions to retain seemingly unprofitable customers based on real options theory and
calculating the CLV.
In previous studies, the relationship between omnichannel marketing strategy and other
marketing topics such as word of mouth, customer buying intentions, customer commitment,
perceptual quality of service, perceptual risk and customer acceptance of online shopping,
customer loyalty, sales trust, etc., has been investigated.
On the other hand, the effect of this strategy on other topics of marketing science such as
customer lifetime value, customer perceptual value, strengthening the brand equity of the
organization, etc., has been less studied. Therefore, in this study, through an emphasis on
omnichannel banking customers, the effect of omnichannel marketing on customer lifetime
value has been investigated.
The research objectives of this case study on Agriculture Bank of Iran included identifying
the behavior of omnichannel customers in terms of the sequence of using banking channels
and comparing their lifetime value with each other and with other customers to identify the
effect of omnichannel marketing strategy on CLV and identifying the optimal behaviors of
omnichannel customers to enhance their CLV. In this regard, the following hypotheses are
presented:
Hypothesis 1): There is a significant difference between omnichannel customers’
lifetime value and other customers’.
Hypothesis 2): There is a significant difference between omnichannel customers’
lifetime value that have different behaviors.
The reason why Agriculture Bank of Iran has been selected as the research area is that in
addition to selecting its target customers from the agricultural, livestock, and food industries
in different geographical locations, this bank also emphasizes on financing from other sectors
and industries. This diversity in its target market has led to the diversity of its customers, and
this has made it necessary to identify the CLV and their classification and plan to improve the
CLV of each type of customer classes.
Furthermore, this bank uses various new banking systems such as ATMs, telephone banks,
Internet banks, mobile banks, etc., in order to provide banking services to customers. This
issue has led to diversity, breadth, and non-integration in banking channels. As a result, it is
necessary to pay attention to the omnichannel strategy.
On the other hand, this bank is a state-owned bank and depends on the strategies and
policies of higher institutions and organizations in formulating and implementing its
marketing strategies and programs; for this reason, less attention has been paid to new
marketing principles such as CLV and strategies such as omnichannel marketing.
Therefore, conducting this study can help this bank succeed in a competitive market.

4. Research Methodology

This study is exploratory, applied, and sectional in terms of approach, purpose, and horizon,
respectively. It is mixed (qualitative-quantitative) in terms of data collection method; in the
qualitative stage, interviews with experts and in the quantitative stage, the survey of real data
and a questionnaire were used for data collection.
56 Rezaei et al.

To achieve research purposes in the quantitative stage, the random sampling method was
used to extract a dataset of two-year banking transactions (2018 and 2019) of 138085
customers of Agriculture Bank of Iran, which were performed using different banking
channels. Then, using SPSS Clementine 12 software and through the association rule mining
method, the rules in the omnichannel customer transaction data set were identified and
customers were classified according to the identified rules.
Omnichannel customers are customers who follow the recognized association rules and
therefore use different banking channels to receive banking services, e.g., customers who first
use ATM, then mobile bank, and finally bank branches to complete the process of receiving
their banking services. Naturally, other customers who do not use the identified association
rules are non-omnichannel customers.
Next, omnichannel customers who followed association rules were categorized into
different groups.
In the qualitative stage, interviews (with 40 experts) and theoretical saturation were used to
identify CLV indicators. In addition, to identify the significance coefficients of indicators
using AHP method, a questionnaire was designed by researchers for pairwise comparisons of
indicators. Then, the reliability and validity of the questionnaire were confirmed using the
Kuder-Richardson method and based on the opinions of experts (10 university professors and
bank managers) who were selected using random sampling method. The results were used to
identify the significance coefficients of the indicators.
Then, in quantitative step, customers’ lifetime value was calculated using the identified
indicators (i.e., recency of deposit, deposit consistency, frequency of received services,
frequency of depositing, average deposit amount, and average received credit). CLV of
omnichannel customers’ lifetime value and other customers’ were compared using t-test.
Subsequently, Omnichannel customers were categorized based on the association rules they
follow, and the CLV of omnichannel customers in different categories were compared using
ANOVA method and Tamhan test.
It is worth mentioning that the customers’ banking transactions data in banks are
confidential and customers should be sure about the security of their banking data. Based on
this, banking transaction data in this study was provided to researchers anonymously and it
was not possible to identify customers related to each banking transaction. Therefore, the
principle of fiduciary duty has been observed by banks and researchers.

5. Research Findings

In the following, each of the steps mentioned in Table1 is described and the results are stated.

Table 1. Research Phases

Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68 57

Step 1: Creating the Data Set

The data used in this paper include banking transactions of 138085 customers of Agricultural
Bank of Iran in the period of 2018 and 2019. These transactions include depositing and
withdrawing bank deposits by customers through various banking channels including POS,
ATM, Branch, Internet bank, and Mobile bank.
At the time of writing, the number of active customers of the Agricultural Bank of Iran was
9337999, and according to Morgan’s table, the appropriate number of sample participants was
138085.
In creating this data set, random sampling method was used and customers whose bank
account was active and who used different banking channels to receive banking services were
considered. While some customers have multiple accounts, the bank transactions of these
customers in different accounts were analyzed in a consolidated manner. Therefore, in this
study, each sample referred to a customer who might have one or several accounts.

Step 2: Association Rule Mining

The data were studied based on the Association Rule Mining method using SPSS Clementine
12 software and the existing rules regarding the sequence of using different channels by these
customers were identified.
Then, rules with support and confidence indicators above 50% (determined by researchers)
were considered.
Association rules in the process of using omnichannel customers of Agricultural Bank of
Iran were identified. The specifications of these are given in Table 2.

Table 2. Association Rules for Data Mining

RuleID Antecedent(A) Consequent(B) Support% Confidence% Lift
1 POS* ATM** 70.49 95.88 1.171
2 ATM POS 81.88 82.55 1.171
3 Branch ATM 63.67 74.03 0.904
4 ATM Branch 81.88 57.57 0.904
5 POS Branch 70.49 62.36 0.979
6 Branch POS 63.67 69.03 0.979
7 POS ATM and Branch 70.49 60.18 1.277
8 POS and ATM Branch 67.59 62.76 0.989
9 Branch POS and ATM 63.67 66.63 0.989
10 POS and Branch ATM 53.96 96.51 1.179
11 ATM Mobile Bank 81.88 51.81 1.179
Source: Research findings
*POS (Point of Sale) is a machine that is installed in sales centers and the customer pays his/her purchase price by swiping
his/her debit or credit card in it.
**An Automated Teller Machine (ATM) is an electronic banking outlet that allows customers to complete basic transactions
without the aid of a branch representative or teller. Anyone with a credit card or debit card can access cash at most ATMs.

In Table 2, support is calculated using the equation P(A ∪ B), confidence is calculated
using Eq.1, and lift is calculated using Eq.2.
The lift indicator in rules 1, 2, 7, 10, and 11 is greater than 1, so the relationship between
antecedent and consequent in these rules is a direct relationship.
According to Table 2, based on the association rules, omnichannel customers of
Agricultural Bank of Iran have the following characteristics in the process of receiving
banking services, in terms of using multiple channels:
1. Rule 1: with a probability of 95.88%, customers who use the POS channel will also
use the ATM channel.
58 Rezaei et al.

2. Rule 2: with a probability of 82.55%, customers who use the ATM channel will also
use the POS channel.
3. Rule 7: with a probability of 60.18%, customers who use the POS channel will also
use the ATM and Branch channels.
4. Rule 10: with a probability of 96.51%, customers who use the POS and Branch
channels will also use the ATM channel.
5. Rule 11: with a probability of 51.81%, customers who use the ATM channel will
also use the mobile bank channel.
The lift index in rules 3, 4, 5, 6, 8, and 9 is less than 1; therefore, the relationship between
antecedent and consequent in these rules is an inverse relationship:
6. Rule 3: with a probability of 74.03%, customers who receive banking services from
the Branch channel will not use the ATM channel.
7. Rule 4: with a probability of 57.57%, customers who use the ATM will not use the
Branch channel.
8. Rule 5: with a probability of 62.36%, customers who use the POS channel will not
use the Branch channel.
9. Rule 6: with a probability of 69.03%, customers who receive banking services from
the Branch channel will not use the POS channel.
10. Rule 8: with a probability of 62.76%, customers who use the POS and ATM
channels will not use the Branch channel.
11. Rule 9: with a probability of 66.63%, customers who use Branch channel will not
use POS and ATM channels at the same time.
After identifying the mentioned rules, the sample customers who followed the identified
rules and showed the rules in their purchase behavior were categorized in 11 groups (as Table
3 shows). However, in the next steps, only categories with lift indicator values larger than 1
and direct antecedent-consequent relationship (i.e., classes 1, 2, 7, 10, and 11) were
considered:

Table 3. Customer Classes

Rules ID 1 2 3 4 5 6 7 8 9 10 11
Categories 1 2 3 4 5 6 7 8 9 10 11
number of customers per
3401 3951 3072 3951 3401 3072 3401 3261 3072 2121 3951
category
Source: Research findings

Customer clustering based on identified rules allows customers who have similar behavior
in terms of usage and sequence of channel usage to be placed on the same cluster. Clustering
customers based on their actual behavior will make the results of the analysis more valid.

Step 3: Calculating the CLV

In order to use the results of the Association Rules Mining in effective omnichannel
marketing planning, it is necessary to consider the impact of the identified rules on CLV
improvement. For this purpose, first, it was necessary to identify the CLV indicators of
Agricultural Bank of Iran customers and then determine the coefficients of importance of each
of them. For this purpose, a portfolio of CLV calculation indicators was identified based on
interviews with experts (Agricultural Bank of Iran managers, including 20 heads of branches,
14 deputies of general departments, and 6 heads of general departments) and saturated
sampling method.
Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68 59

In order to conduct pairwise comparison and to analyze and identify the coefficients of
importance of the indicators using AHP method, a questionnaire was prepared by researchers.
The reliability of the questionnaire was examined using Kuder-Richardson model, and based
on the obtained reliability coefficient (0.72), the reliability of the questionnaire was approved.
The validity of the questionnaire was also confirmed by experts (a number of university
professors and bank managers). The decision hierarchy tree of CLV calculation indicators is
as Figure 1.

CLV indicators

Frequency of Recency of deposit

depositing (F2) Frequency of Deposit (R1)
received services consistency ratio
(F1) (R2)
Average of received Average of deposit
credit (M2) amount (M1)

Figure 1. Hierarchical Decision Tree of AHP Model (Source: Research Findings)

Then, 330 managers of the Agricultural Bank of Iran (including 45 heads of departments,
75 deputies of departments, and 210 heads of branches) were selected as samples based on
stratified random sampling and their opinions on the comparison of the importance of CLV
indicators were collected using a questionnaire. Given that the number of managers of the
Agricultural Bank of Iran, as a target community, is equal to 2098 people, according to
Morgan’s table, the sample size was reasonable.
In order to use the AHP model to analyze the collected data, the pairwise consistency rate
was first measured. These calculations showed that the consistency rate was 0.07 and
acceptable (less than 0.1). After that, the geometric mean of the respondents’ opinions about
the questionnaire was calculated and in this way, their opinions were combined. Then, to
determine the priority coefficient of each indicator, first the integrated matrix was normalized
by dividing each of the combined matrix processes on the total column, and then using the
geometric mean of each of the normalized matrix lines, the priority coefficient was identified.
Finally, by dividing each of the coefficients by the sum of the priority coefficients of the
indicators, the calculated coefficients were normalized. Figure 5 shows the average values of
normal pairwise comparisons and the priority coefficients of the indicators.
The normalized matrix and the normalized priority coefficients CLV indicators are listed in
Table 4.

Table 4. The Average of the Normalized Pairwise Comparisons and the Priority Coefficients of the
Indicators
Geometric Normalized Priority
)R1( )R2( )F1( )F2( )M1( )M2(
Mean )Wi( Coefficients
)R1( Recency of deposit 0.06 0.15 0.18 0.01 0.07 0.21 0.08 0.09
)R2( Deposit consistency ratio 0.02 0.05 0.18 0.09 0.09 0.21 0.08 0.09
)F1( Frequency of received services 0.01 0.01 0.04 0.02 0.07 0.04 0.02 0.03
)F2( Frequency of deposit 0.39 0.34 0.26 0.09 0.08 0.13 0.18 0.20
)M1(Average of received credit 0.51 0.44 0.33 0.78 0.62 0.38 0.49 0.56
)M2( Average of deposit rate 0.01 0.01 0.01 0.02 0.07 0.04 0.02 0.02
Source: Research findings
60 Rezaei et al.

Finally, the formula for calculating CLV, which is derived from the RFM model, was
designed as Eq. 3:
CLV=w1R1 + w2R2 + w3F1+ w4F2+ w5M1+w6M2 (3)
w1, w2, w3, w4, w5, w6: in order, priority coefficients recency of deposit (R1), deposit
consistency (R2), frequency of received services (F1), frequency of depositing (F2), average of
deposit amount (M1), and the average of received credit (M2) which are equal to the Wi
column in Table 4. Therefore, the CLV formula of the Agricultural Bank of Iran customers is
equal to Eq. 4:
CLV=0.09R1 + 0.09R2 + 0.03F1+ 0.2F2+ 0.56M1+0.02M2 (4)
Then, the CLV of the sample customers was calculated using the above formula and based
on the information in the dataset transaction of the customers.

Step 4: Comparison of the CLV of Omnichannel and Other Customers

Here, omnichannel customers were customers who used the banking channels based on the
association rules identified in Figure 6, and their lift indicator was greater than 1, i.e., the
customers of categories no. 1, 2, 7, 10 and 11, which are categorized in group 1 in Table 5.
Other customers are customers whose use of different banking channels does not have a
specific association rule; these customers are categorized in group 2. This group included
customers who were in dataset, but during the time considered (2018 and 2019) did not use
any banking channels have not been considered; the number of these customers is 105,181.
To compare the CLV of these two groups, t-test was used. H0 and H1 of this test would be
as Eq. 5:
H0: µ1= µ2
H1: µ1≠ µ2
(5)
µ1: CLV average of omnichannel customers
µ2: CLV average of other customers
The outputs of SPSS 22 for t-test are showed in Table 5 and 6.
Table 5. T-Test Results
Group statistics
Groups N Mean Std. Deviation Std. Error Mean
1 11738 301.74 15.88 .146
CLV
2 21166 128.42 90.59 .622
Resource: Research findings

The number of omnichannel customers (group 1) is 11738, their CLV average is 301.74,
and their standard deviation is 15.88. The number of other customers (group 2) is 21166, their
CLV average is 128.42, and their standard deviation is 90.59.
Table 6. Independent Samples Test
Levene's test for
equality of t-test for equality of means
variances
95% confidence interval
Mean Std. error of the difference
f sig. t df Sig. (2-tailed)
difference difference
Lower Upper
Equal variances
79738 .706 205 32902 .006 173.31 .84 171.66 174.96
assumed
CLV
Equal variances
270 23446 .006 173.31 .63 172.06 174.56
not assumed
Source: Research findings
Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68 61

In Table 6, since the significance value is greater than 0.05, the assumption of the
equivalence of variances is accepted and first-line data are used to interpret the table content.
Since the test is two-tailed and the significance value of the first row is equal to 0.006, the H0
– which shows that the average CLV of the two groups is equal – is not acceptable at 0.95
confidence level.
Therefore, based on the analysis of Table 6, research hypothesis 1 is confirmed, and we
might maintain with 95% confidence that there is a significant difference between
Agricultural Bank or Iran omnichannel customers’ lifetime value and those of its other
customers. Based on the findings, the amount of this difference is equal to 173.31 (134%).

Step 5: Comparison of the CLV of Omnichannel Customers

In the fifth step, using SPSS 22 software, the average CLVs of omnichannel customers were
compared and their differences were examined, as shown in Table 7.

Table 7. ANOVA Results

Sum of Squares df Mean Square F Sig.
Between groups 345502 4 86375 387 .000
Within groups 2615693.07 11733 222
Total 2961195 11737
Source: Research findings

The significance value shown in Table 7 is 0.000. Therefore, hypothesis 2 is confirmed,

and we might maintain with 95% confidence that there is a significant difference between
omnichannel customers’ lifetime values that have different behaviors.
Having different behaviors mean they follow the different association rules identified in
the use of different banking channels.
Then, in order to achieve the research purposes and identify the optimal behavior
(following the optimal association rules) of omnichannel customers to improve the customers’
lifetime values, the average CLV of omnichannel customers were compared.
Selecting the test type to compare the average CLV of omnichannel customers in different
categories depends on the result of variance homogeneity test. In this study, Levene’s test was
used for this purpose.

Table 8. Test of Homogeneity of Variances

Levene statistic df1 df2 Sig.
296 4 11733 .000
Source: Research findings

Based on Table 8, because significance is less than 0.05, it can be concluded that there is a
significant difference among the variance of CLVs of omnichannel customers in different
categories; therefore, the Tamhane’s test, in which the variance of the groups is assumed
heterogeneous, was used to compare the average CLVs of omnichannel customers in different
categories.
62 Rezaei et al.

Table 9. Multiple Comparisons Between Categories Based on CLV

95% Confidence interval
Category(I) Category(J) Mean difference (I-J) Std. error Sig.
Lower bound Upper bound
2 2.82* .14 .000 2.40 3.23
7 6.23* .28 .000 5.45 7.02
1 10 -3.08* .80 .001 -5.32 -.84
11 -16.42* .40 .000 -17.57 -15.28
1 -2.82* .14 .000 -3.23 -2.40
7 3.41* .29 .000 2.58 4.24
2 10 -5.90* .80 .000 -8.16 -3.64
11 -19.25* .41 .000 -20.42 -18.07
1 -6.23* .28 .000 -7.024 -5.45
2 -3.41* .29 .000 -4.24 -2.58
7 10 -9.32* .84 .000 -11.67 -6.96
11 -22.66* .48 .000 -24.01 -21.31
1 3.08* .80 .001 .84 5.32
2 5.90* .80 .000 3.64 8.16
10 7 9.32* .84 .000 6.96 11.67
11 -13.34* .89 .000 -15.84 -10.84
1 16.42* .40 .000 15.28 17.57
2 19.25* .41 .000 18.07 20.42
11 7 22.66* .48 .000 21.31 24.01
10 13.34* .89 .000 10.84 15.84
Source: Research finding *. The mean difference is significant at the 0.05 level.

Table 10. Descriptive Table of Comparisons Between Different Categories

Source: Research findings

Based on the results presented in tables 9 and 10, which are the result of the Tamhane test,
we might maintain with 95% confidence that there is a significant difference among the
omnichannel customers’ lifetime value in different categories and their average CLVs. The
values (ordered from more to less) in categories 11, 10, 1, 2, and 7 are equal to 318.45,
305.11, 302.02, 299.20, and 295.79, respectively.
The differences among the average CLVs of omnichannel customers in different categories
are shown in Table 11.

Table11. Comparison the CLV of Omnichannel Customers in Different Categories

Source: Research findings

Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68 63

6. Discussion and Conclusions

In today’s competitive market, where the life cycle of products and services and the
sustainability of companies’ relative benefits are shortened, it is not possible to succeed
without studying customer behavior and developing strategies to enhance their lifetime
values. Delivering banking services through numerous and diverse channels, developing an
omnichannel strategy, and coordinating and integrating the different channels have made it
more necessary to shape customers’ behavior and enhance their lifetime values.
Studying the big data of how to use banking services and identifying the association rules
in the behavior of banking customers can be effective in implementing the omnichannel
strategy and integrating between banking services and channels.
In a study by Aggelis (2004) the way of discovery of association rules between different
types of e-banking payment offered by a bank is described along with experimental results.
The basic outcome is that value added tax (VAT) payment orders and social insurance
institute (SII) payment orders are the most popular and interconnected strongly. Kargari &
Eshraghi (2018) studied the fraud behavior in banking service using association rule data
mining; Findings suggest that the employment of both rule-based and clustering-based
components leads to the detection of more frauds while fewer alarms will go off.
In these studies, or similar studies that have been reviewed in this paper, the association
rules of consumer behavior in receiving banking services have been studied and the channels
of providing these services have not been examined. While this paper examines the
association rules of using banking channels, using data mining.
In the present study, which was conducted as a case study in the banking industry on the
Agriculture Bank of Iran, first the current behaviors of sample customers of the Agricultural
Bank of Iran in using different banking channels were identified through the Association Rule
Data Mining method. A summary of results provided in the previous section of this article is
shown in Figure 2.

Lift>1 Lift<1
Figure 2. Diagram of Association Rules Mining Result (Source: Research Finding)
64 Rezaei et al.

In association rules with a lift greater than 1, i.e., rules in which the relationship between
antecedent and consequent is direct, POS plays a key role. These association rules, with
different probabilities, indicate that customers who use POS will also use ATM and customers
who use ATM will also use POS. Customers who use ATM tend to use Mobile Bank.
Customers using POS tend to use ATM and Branch at the same time. Customers who use
POS and Branch simultaneously also tend to use ATMs.
Among the association rules whose lift is less than 1, i.e., rules in which the relationship
between antecedent and consequent is inverse, the use of Branch plays a special role, so that,
customers who use bank branches don’t tend to use ATM, POS, and POS & ATM
(simultaneously). In addition, customers who use POS or ATM or POS & ATM
(simultaneously) will not be willing to use Branch.
These association rules can be the basis of omnichannel marketing, so the channels of
these association rules should be integrated and omnichannel customers should be directed to
these integrated channels.
Then, sample customers were categorized based on the use of identified association rules.
In order to calculate the CLV of Agricultural Bank of Iran customers, first the indicators
measuring CLV and their priority coefficients were identified based on the opinion of bank
experts and using AHP method.
Recency of deposit, deposit consistency, frequency of received services, frequency of
depositing, average deposit amount, and average of received credit were the main indicators
of Agricultural Bank customers’ lifetime value. Due to the similarity of banking services and
customers of banks in Iran, these indicators can be generalized to other banks.
Among these indicators, recency of deposit and average of deposit amount are more
important than others; accordingly, banks should try to strengthen these indicators to improve
their customers’ lifetime value.
In the next step, the CLV of customers was calculated using identified indicators and two-
year (2018 and 2019) history of sample customers’ banking transactions.
Subsequently, the CLV average of omnichannel customers (customers who follow
identified association rules) and other customers was compared using t-test. The results
showed that with 95% confidence, the difference between the omnichannel customers’
lifetime value and the other customers’ was significant and the CLV average of omnichannel
customers was 2.34 times of the CLV of other customers. So, Hypothesis 1 was confirmed.
The priority of association rules identified in omnichannel marketing management required
comparing the CLV average of omnichannel customers, who use different association rules.
Therefore, first, the sample customers were categorized based on the identified association
rules, and the difference between the omnichannel customers’ lifetime value were examined
using ANOVA method as well as Tamhane method. The results showed that with 95%
confidence, the difference between CLV of omichannel customers of different categories was
significant (Hypothesis 2 was confirmed) and the average CLV of omnichannel customers of
the Agricultural Bank of Iran (ordered from more to less) in categories 11 (ATM →
MobileBank), 10 ( POS & Branch → ATM), 1 (POS → ATM), 2 (ATM → POS) and 7 (POS
→ ATM & Branch) were 305.11, 318.45, 302.02, 299.20, and 295.79, respectively. Table 11
showed the differences between the average CLV of omnichannel customers of different
categories.
Based on these findings, the implementation of omnichannel strategy enhances customers’
lifetime value and omnichannel customers’ lifetime value is more than others.’ This
highlights the importance of an omnichannel marketing strategy in enhancing customers’
lifetime value.
Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68 65

On the other hand, the behavior of omnichannel customers differs significantly based on
the sequence of the use of channels, and the CLV of customers who use the 11, 10 and 1
association rules have the highest CLV. Therefore, in developing an omnichannel strategy,
special emphasis should be placed on these association rules.

7. Suggestions

The above results indicate that in the competitive market of banking services, where banks
have offered several electronic banking channels to provide services and customers’
knowledge has been upgraded to use electronic banking channels, adopting a omnichannel
marketing strategy to expand visibility to channels and create integration between them can
lead to increased customer lifetime value (CLV). Therefore, the following points are
suggested.
1. Identify customer association rules through a data warehousing system: Changes in
customer behavior are becoming faster day by day, so the identification of association rules in
the behavior of omnichannel customers should be done continuously. It is therefore
recommended to the Agricultural Bank of Iran to develop and use systems such as data
warehouse to analyze customer behavior, discover and categorize customer needs and
interests, and use the results to personalize services, which is one of the goals of omnichannel
marketing.
2. Develop banking channels based on effective association rules: In order to improve
its omnichannel customers’ lifetime value, the Agricultural Bank of Iran should develop their
access to POS, ATM, and mobile banking channels, which are emphasized in the recognized
association rules; the use of ATM channels will encourage customers to use other electronic
banking channels and reduce the use of the branches.
3. Reengineer the process of banking services based on effective association rules: The
process of providing banking services to customers might be designed in such a way that
there is communication and integration between different channels based on the findings of
association rule mining, and customers can access the history and results of services when
using different channels and can continue the process of receiving services from one channel
to another. For example, it is suggested that the transaction history of the customers in any
channel can be accessed on the others, or that it can be possible in mobile bank for customers
to verify transactions such as the payment of checks issued by them and the payment required
to their verification.
4. Homogenize the channels: Different channels, which are in sequence according to the
identified association rules, should be homogeneous in terms of content, color, design,
usability, usability, and process;
This homogeneity can lead to the homogeneity of customers’ perceptions about different
channels, which is one of the important goals of omnichannel marketing.
Theoretically, in order to conduct future studies, the following points are suggested.
1. Develop a comprehensive model: In studies on omnichannel marketing, the various
features, dimensions, and structures of omnichannel marketing have been studied separately
and a comprehensive model has not been designed so far. However, success in designing an
effective omnichannel strategy requires a comprehensive understanding of the various
dimensions of the omnichannel concept. It is then suggested that the identification of different
dimensions of omnichannels and the design of a comprehensive model for it be considered in
future research.
2. Conduct an interdisciplinary study: Studies on the omnichannel strategy can be
divided into two areas, namely information technology and marketing. Studies in the field of
66 Rezaei et al.

information technology have dealt only with the hardware and software dimensions, and
studies in the field of marketing have dealt only with the communication, advertising, sales,
and distribution dimensions.
The successful development and implementation of an omnichannel strategy requires
coordination between these disciplines, which can be considered in future research with a
team of IT, marketing, and banking researchers.
3. Conduct this study in other industries: The present study has been conducted in the
banking industry and specifically in the Agriculture Bank of Iran. It is suggested that this
research might be done in other industries as well.

8. Limitations

Researchers faced some limitations in conducting this research, including what follows.
1. Conducting this research using data mining method required access to real data of
customers’ banking transactions and identification of their behavior. Due to the
confidentiality of customers’ banking transactions, this access was difficult and coded and
without the possibility of identifying customers, while access to some customer demographics
could increase the validity of the research findings.
2. Considering that this research has been conducted in the banking industry and
specifically as a case study in the Agricultural Bank of Iran, the results are less generalizable
to other industries. Of course, the results can be used to conduct research in other industries.
Iranian Journal of Management Studies (IJMS) 2022, 15(1): 49-68 67

Reerences

Bhatnagar, A., & Syam, S. (2014). Allocating a hybrid retailer’s assortment across retail stores:
Bricks-and-mortar vs online. Journal of Business Research, 67(June), 1293-1302.
Cao, L., & Li, L. (2015). The impact of cross-channel integration on retailers’ sales growth. Journal of
Retailing, 91(2), 198–216.
Chen, Y., Cheung, C. M. K., & Tan, C. W. (2018). Omnichannel business research: Opportunities and
challenges. Decision Support Systems, 109(May), 1-4.
Chopra, S. (2016). How omnichannel can be the future of retailing. Decision, 43(2), 135–144.
Cummins, S., Peltier, J. W., & Dixon, A. (2016). Omnichannel research framework in the context of
personal selling and sales management. Journal of Research in Interactive Marketing, 10(1), 2–
16.
Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behavior: An introduction to theory
and research. Reading, MA: Addison-Wesley
Gallino, S., & Moreno, A. (2014). Integration of online and offline channels in retail: The impact of
sharing reliable inventory availability information. Management Science, 60(6), 1434–1451.
Gao, F., & Su, X. (2016). Online and offline information for omnichannel retailing. Manufacturing
Service & Operations Management, 19(1), 84–98.
Gupta, S., Hanssens, D., Hardie, B., Kahn, W., Kumar, V., Lin, N. (2006). Modeling customer lifetime
value. Journal of Service Research, 9(2), 139–155
Herhausen, D., Binder, J., Schoegel, M., & Herrmann, A. (2015). Integrating bricks with clicks:
Retailer level and channel level outcomes of online-offline channel integration. Journal of
Retailing, 91(2), 309–325.
Hsia, T. L., Wu, J. H., Xu, X., Li, Q., Peng, L., & Robinson, S. (2020). Omnichannel retailing: The
role of situational involvement in facilitating consumer experiences. Information &
Management, 57(8), 103390.
Hure, H., Picot-Coupey, K., & Ackermann. C.L. (2017) Understanding omnichannel shopping value:
A mixed-method study. Journal of Retailing and Consumer Services, 39 (November), 314-330.
Kim, J. C., & Chun, S. H. (2018). Cannibalization and competition effects on a manufacturer’s retail
channel strategies: Implications on an omni-channel business model. Decision Support System,
109(January), 5–14.
Lee, Z., Chan, T. K. H., Yee-Loong C.A., & Thadani, D. R. (2019). Customer engagement through
omnichannel retailing: The effects of channel integration quality. Industrial Marketing
Management, 77(February), 90-101.
Moliner, J. R. S., & Tena, M. A. M. (2016). Customer equity and clv in Spanish telecommunication
services. Journal of Business Research, 69(10), 4694-4705.
Nikkhahan, B., Badrabadi, A. H., & Tarokh, M. J. (2011). Customer lifetime value model in an online
toy store. Journal of Industrial Engineering International, 7(12), 19–31.
Saghiri, S., Wilding, R., Mena, C., & Bourlakis, M. (2017). Toward a three-dimensional framework
for omni-channel. Journal of Business Research, 77(August) 53–67.
Shao, X. (2021). Omnichannel retail move in a dual-channel supply chain. European Journal of
Operational Research, 294 (3), 936-950.
Shen, X.L., Li, Y.J., Sun, Y., & Wang, N. (2018). Channel integration quality, perceived fluency and
omnichannel service usage: The moderating roles of internal and external usage experience.
Decision Support Systems, 109(May), 61-73.
Shi, Y., Wu, B., Chen, N., Chen, A., Li, J., & Li, H. (2019). Determination of effective management
strategies for scenic area emergencies using association rule mining. International Journal
Disaster Risk Reduction. 39, 101208.
Suárez, M. M., & Tejero, N. C. (2021). Why do banks retain unprofitable customers? A customer
lifetime value real options approach. Journal of Business Research, 122, 624-626
Tarokh, M. J., & Esmaeili Gookeh, M. (2019). Modeling patient’s value using a stochastic approach:
An empirical study in the medical industry. Computer Methods and Programs in Biomedicine,
176(July), 51-59.
68 Rezaei et al.

Valenzuela, L., Torres, E., Hidalgo, P., & Farias, P. (2014). Salesperson CLV orientation’s effect on
performance. Journal of Business Research, 67(December), 550-557
Van Baal, S. (2014). Should retailers harmonize marketing variables across their distribution
channels? An investigation of cross-channel effects in multi-channel retailing. Journal of
Retailing and Consumer Services, 21(6), 1038–1046.
Verhoef, P. C., Kannan, P. K., & Jeffrey Inman, J. (2015). From multi-channel retailing to omni-
channel retailing introduction to the special issue on multi-channel retailing, Journal of
Retailing, 91(2), 174–181.
Wagner, G., Schramm-Klein, H., & Steinmann, S. (2018). Online retailing across e-channels and e-
channel touchpoints: Empirical studies of consumer behavior in the multichannel e-commerce
environment. Journal of Business Research, 107 (November), 256-270.
Wang, F., Liming, L., Kangping, L., Neven, D., Miadreza, S. K., & Joao P. S. C. (2019). Impact
factors analysis on the probability characterized effects of time of use demand response tariffs
using association rule mining method. Energy Conversion and Management, 197(October),
111891.
Aggelis, V., (2004). Data mining V, UK, Wessex Institute of Technology
Kargari, M., & Eshghi, A. (2018). A Model Based on Clustering and Association Rules for Detection
of Fraud in Banking Transactions. Proceedings of the 4th World Congress on Electrical
Engineering and Computer Systems and Sciences, Madrid, Spain, August 21–23