A Novel Approach for the Customer Segmentation Using Clustering Through Self-Organizing Map

A Novel Approach for the Customer Segmentation Using Clustering Through Self-Organizing Map

Debaditya Barman (Visva-Bharati University, Santiniketan, India) and Nirmalya Chowdhury (Jadavpur University, Kolkata, India)
Copyright: © 2019 |Pages: 23
DOI: 10.4018/IJBAN.2019040102


Customer segmentation is the process of forming smaller groups of customers according to their characteristics. Now companies can develop proper marketing strategies for each group to get the desired results. This type of direct marketing is practiced by most organizations from the size of smallest start-up to the Fortune 500 leaders. Clustering is the ideal data mining technique for customer segmentation. In this article, the authors have proposed a clustering algorithm based on the self-organizing map and minimum spanning tree for customer segmentation. The authors have used several synthetic and real-life datasets to evaluate the clustering performance of their approach. To demonstrate the effectiveness of the authors' proposed approach, they have trained few classifiers with the groups extracted from a direct marketing campaign of a Portuguese banking institution and show that the classification accuracy is better compared to the results obtained in some previous work where the full dataset has been used to train the same classifiers.
Article Preview

1. Introduction

Customer segmentation plays a crucial role when a company is trying to reach out to a target market segment. This enables marketing teams to optimize their budget in obtaining the best possible results in terms of sell of a product. After customer segmentation is properly implemented marketers can create specific messages for each target market segment. This type of communication will help the marketers to build rapport with the target audiences. Perhaps the most common ways to segment the market are demographic, behavioral, geographic, and psychographic in nature (Elrod & Winer, 1982). Demographic segmentation can be done using the attributes like age, gender, income, job, education, religion, race, nationality, marital status etc. For behavioral segmentation we have to look for the attributes of purchase patterns, loyalty status, readiness state, attitude toward product etc. Geographic segmentation deals with the attributes like region, country size, city size, population density, climate etc. Psychographic segmentation considers attributes like social class, lifestyle, personality etc. In this work, we have used customer’s demographic information for segmentation. Although customer segmentation divides the entire customer database into a number of groups, the size of each such group is still considerably large. One may imagine a scenario where a company is trying to market a product with tight marketing budget. It may be noted that even if the entire customer database is segmented into a number of groups the marketing budget may not be sufficient enough to reach all the customers of any such group. With this limitation in mind we are proposed to use suitable classifiers to make ranking of customers within each such said groups in the order of their possibilities of being influenced by the advertisement for a specific product.

Clustering is a very popular data mining technique that can be used for customer segmentation process. Clustering is an unsupervised method for grouping similar type objects from the heterogeneous collection of data points and discovering the inherent structure present in the data set that plays an important role in data mining (Plamondon & Srihari, 2000). For a given data set IJBAN.2019040102.m01, what one perceives to be the groups present in IJBAN.2019040102.m02 by viewing the scatter diagram of IJBAN.2019040102.m03, is termed as natural groups of IJBAN.2019040102.m04. Ben-Hur (Ben-Hur, Elisseeff, & Guyon, 2001) proposed a method to detect the clusters present in the data set based on “natural grouping.”

Let the set of IJBAN.2019040102.m05 patterns IJBAN.2019040102.m06 and if it consists IJBAN.2019040102.m07 clusters IJBAN.2019040102.m08 then the following conditions should satisfy:

  • 1.

    IJBAN.2019040102.m09, for IJBAN.2019040102.m10;

  • 2.

    IJBAN.2019040102.m11 for IJBAN.2019040102.m12; and

  • 3.

    IJBAN.2019040102.m13, where IJBAN.2019040102.m14 represents Null set.

Complete Article List

Search this Journal:
Open Access Articles
Volume 8: 4 Issues (2021): Forthcoming, Available for Pre-Order
Volume 7: 4 Issues (2020)
Volume 6: 4 Issues (2019)
Volume 5: 4 Issues (2018)
Volume 4: 4 Issues (2017)
Volume 3: 4 Issues (2016)
Volume 2: 4 Issues (2015)
Volume 1: 4 Issues (2014)
View Complete Journal Contents Listing