Two-tiered Clustering Classification Experiments for Market Segmentation of EFTPOS Retailers

Abstract

Almost all the papers reported in the literatures on market segmentation modeling using retail transaction data deal with finding groupings of customers. This paper proposes the application of clustering and classification techniques for finding groupings of retailers who use the Electronic Funds Transfer at Point of Sale (EFTPOS) facilities of a major bank in Australia in their businesses. The RFM (Recency, Frequency, Monetary value) analysis on each retailer is used to reduce the large data set of customer purchases through the EFTPOS network into attributes that may explain the business activities of the retailers. Preliminary results published in Singh, show that groupings of retailers with distinct combinations of RFM values can be established. Encouraged by the promising business insights gained from the clustering experiments using the RFM values, in this paper, we report our findings in extracting business rules pertinent to each cluster. For this purpose, we incorporate attributes of the EFTPOS transaction data in addition to the derived RFM attributes to build a decision tree to facilitate the extraction of the business rules.