Comparative Evaluation of SOM-Ward Clustering and Decision Tree for Conducting Customer-Portfolio Analysis

Abstract

Analyzing customer’s base for the purpose of retaining and attracting the most valuable customers still stands the main problem facing companies in this modern age. The process of conducting customer portfolio analysis (CPA) makes most existing customer relationship management systems (CRMS) lack ability to extract hidden information and knowledge from pool of data stored in customer databases or data warehouses, to conduct market segmentation due to the implementation techniques used. In this paper, a two-level hybrid approach that combines Self-Organizing Maps-Ward’s clustering denoted as (SOM-Ward) and decision trees data mining technique is proposed. The dataset used in this study were acquired from the loyalty card system, containing 1,480,662 customers, and sales information from several department stores. SOM-Ward clustering was used to conduct customer segmentation, by dividing the customer base into distinct segment of customers with similar characteristics and behavior. A Decision Tree was used to partition a large collection of data into smaller sets in order to identify the characteristics that can tell high-spending customers from low-spending ones. The result of the prediction performance of the decision tree shows that the overall accuracy of the model is 72.6%. 82.8% of the high spending customers are correctly classified, while only 62.4% of the low spending customers are correctly classified. This study revealed that the approach that combines SOM-Ward clustering and decision trees data mining technique outperforms SOM-based approach and Decision Trees individually for market segmentation, classification, and data exploration problems. Keywords: Customer relationship management (CRM), customer portfolio analysis (CPA), Self-Organizing Maps (SOM), Ward’s clustering, decision trees.