Presenting and comparing three two-stage models for customer segmentation based on their value using K-Means, SOM and RFM data mining tools (Case study: Iran Apple Center chain stores)

Number of pages: 124 File Format: Not Specified File Code: 29543
Year: Not Specified University Degree: Not Specified Category: Industrial Engineering
Tags/Keywords: Industries
  • Part of the Content
  • Contents & Resources
  • Summary of Presenting and comparing three two-stage models for customer segmentation based on their value using K-Means, SOM and RFM data mining tools (Case study: Iran Apple Center chain stores)

    Dissertation for obtaining a master's degree in industrial engineering

    October 2011

    Identifying the value[1] of customers, one of the main components of success in the store. There are various methods that have received more attention nowadays. Chain stores [2] are in contact with different groups of customers, and due to limited resources, they must rank customers based on their value in order to allocate an appropriate portion of marketing resources to more valuable customers and earn more profit. Therefore, we use data mining techniques [3] to segment customers, a lot of research has been done in this field. Many of these researches from  RFM model [4] has been used for customer segmentation [5]. This model includes three indicators of delay, frequency and monetary value for analyzing the buying behavior of customers and can represent the behavioral value of customers. In this research, a comprehensive methodology including three segmentation models based on the development of RFM and SOM models[6] and K average[7] has been presented, and transaction data and cognitive demographics have also been examined for customer identification. The proposed models were implemented in chain stores of Apple Center of Iran [8] and 347 customers were investigated. For transactional data, transactions recorded in the store's information center were used, and demographic data were also asked from each customer over the phone. These customers were segmented with all three different models, and finally, these models were evaluated and compared with Davis Boldin's index and sum of squared error[9]. According to the Davis Bouldin index, the first model has shown better performance in this case study, but based on the criterion of sum of squared error, the second model has a better performance and this difference is due to the nature of these two criteria.

    Key words: chain store, customer segmentation, customer value, data. Mining, RFM, SOM, K  Mean  

    Introduction

      Changing consumer buying habits and emerging technologies have created a massive upheaval across the retail industry. Consumers are challenging the industry with the way they live today. Supported by emerging technologies, consumers are more focused on price and convenience than ever before. Hence, retailers must be able to clearly differentiate themselves from their competitors with excellent customer service, enabled by technology. It is important to note this to avoid customer churn, as the cost of acquiring new customers is much higher than retaining them. The key to survival in this competitive industry is to understand and know customers better. One of the methods used to understand customers and identify homogeneous groups is customer segmentation. Customer segmentation is a significant issue in today's competitive business environment. Many studies have investigated the application of data mining technology in customer segmentation and have achieved its effects. The data mining method is of great help to researchers to extract knowledge and hidden information from the data. Customer analysis, which is a requirement for segmentation, enables stores to be more in tune with customer behavior. In addition, the planning section can create more clarity in the planning process by highlighting the needs of marketing programs and specific customer groups.

      In the first chapter of this research, the general presentation of the research, statement of the problem, necessity of doing it and research questions will be discussed. The proposed methodology and models are also presented in the majority of the diagram.

      In the second chapter, the research records and theoretical literature of the research are presented. The articles published in the field of segmentation and its literature are examined, and the existing algorithms for customer segmentation and its application in different industries are stated.

      In the third chapter, the research method is proposed.The method of data collection, the statistical population and the sampling method stated at the end, the new models presented have also been explained.

      In the fourth chapter, the numerical results obtained from the implementation of the model in the chain stores of Apple Center of Iran are described and the implementation steps of the proposed models in this chain store are explained.

      In the fifth chapter, the comparison of the models and the results of the research is stated. The questions raised are answered and the feedback from the experts about the results and findings of the research is also presented. Also, the limitations of the research are stated and at the end, suggestions for future research are presented. Chapter 1: Overview of the research. In the first chapter of this research, the general presentation of the research, the statement of the problem and the necessity of doing it, and the research questions will be discussed. The proposed methodology and models are also presented in the majority of the diagram. 1-2 Necessity of conducting research. In every business, companies are in contact with different groups of customers. Therefore, considering the limited resources, they must rank customers based on their value in order to allocate an appropriate portion of marketing resources to more valuable customers and earn more profit.

      Despite this high competition, companies should try to attract new customers and retain more valuable customers with value-added activities. Customer relationship management improves the company's relationship with customers to achieve more profit (Tabai and Fathian [1], 2011). Companies have a lot of valuable information about customers and their past shopping experiences. The use of this information helps them to examine the interests, satisfaction and loyalty of customers. Therefore, by using data mining techniques and segmenting customers into different groups, companies can have profitable marketing strategies. 1-3 Statement of the problem Customer value is an important issue in customer relationship management and there are many ways to find it. In this research, we present a comprehensive methodology including three two-stage models for segmenting customers based on their value. In this methodology, we use two databases, including personal profiles of customers[2] and transactional data[3], as shown in Figure 1-1.

      In the first model of this methodology, we first perform segmentation based on demographic data [4] of customer profiles using a self-organized neural network [5], then we re-segment each of the segments obtained from the first stage based on transaction data (weighted RFM) based on the K average algorithm. We obtain the optimal K in each cluster with the Davies-Bouldin method and finally rank the obtained sections based on their value. which is shown in Figure 1-2.

      In the second model of this methodology, we first classify customers based on transaction data (weighted RFM) using the K average segmentation algorithm. In this method, the optimal K value is determined in advance by the Davies Bouldin index. Then we re-segment each segment obtained from the first stage based on the demographic data using self-organizing neural network and finally we rank the segments based on their value. which is shown in Figure 1-3.

    In the third model of this methodology, firstly, customers are segmented using self-organizing neural network, based on demographic variables and transactional variables (weighted RFM), then the number of clusters obtained (k) and cluster centers as input to the K-means method to re-segment customers based on variables We use cognitive and transactional demographics and finally rank the obtained segments based on their value. which is shown in Figure 1-4.

  • Contents & References of Presenting and comparing three two-stage models for customer segmentation based on their value using K-Means, SOM and RFM data mining tools (Case study: Iran Apple Center chain stores)

    Introduction. 1

    Chapter one: Research overview. 4

    1-1 Introduction. 4

    1-2 Necessity of doing research. 4

    1-3 statement of the problem. 5

    1-4 research questions. 10

    1-5 research methods and tools. 10

    Chapter Two: Theoretical foundations. 12

    2-1 Introduction. 12

    2-2 Customer relationship management. 13

    2-2-1 Advantages of using customer relationship management. 14

    2-2-2 types of customer relationship management technologies. 15

    2-3 data mining. 17

    2-3-1 Data mining processes. 19

    2-3-2 Data mining activities and applications. 20

    2-5 clustering. 25

    2-5-1 Advantages of the clustering method. 26

    2-5-2 K average algorithm. 27

    2-4 division. 29

    2-4-1 Objectives of division. 30

    2-4-2 Advantages of customer segmentation. 31

    2-4-3 general criteria of division. 32

    2-6 RFM algorithm. 34

    2-6-1 Advantages of RFM algorithm. 35

    2-7 self-organizing maps. 35

    2-7-1 Application of self-organizing maps. 36

    2-7-2 Topology of self-organizing maps. 37

    8-2 Application of segmentation in different industries. 40

    2-9 research background. 42

    The third chapter: research method. 56

    3-1 Introduction. 56

    3-2 research method. 56

    3-3 Statistical population and sampling method. 57

    3-4 description of the model. 58

    3-5 model validity evaluation. 64

    Chapter 4: Data analysis (Implementation of the model in Apple Iran Center) 67

    4-1 Introduction. 67

    4-2 understanding and recognition of data 67

    4-3 data preparation 69

    4-4 determining the weight of transactional parameters (RFM) using the hierarchical analysis process. 71

    4-5 Ensuring the correctness of the weights calculated in the AHP technique using the compatibility rate test 74

    4-6 Implementation of the models in Apple Iran Center. 76

    4-6-1 Two-stage segmentation of customers using the first model. 76

    4-6-2 Two-stage segmentation of customers using the second model. 83

    4-6-3 Two-stage segmentation of customers using the third model. 90

    4-7 Evaluation of models 94

    Chapter 5: Conclusion. 97

    5-1 Introduction. 97

    5-2 Summary and conclusion. 97

    5-3 feedback from experts about the results and findings of the research. 99

    4-5 limitations of research. 100

    5-5 suggestions for future research. 100

    Persian sources. 101

    English sources. 102

    Appendixes 109

Presenting and comparing three two-stage models for customer segmentation based on their value using K-Means, SOM and RFM data mining tools (Case study: Iran Apple Center chain stores)