Credit Ranking of Bank Customers (An Integrated Model of RFM, FAHP and K-means)

Abstract

In this paper, with the aim to rank customers in terms of credit, three patterns namely Hsieh (RFM), FAHP, and K-means were integrated. The main effective factors on ranking customers including transactions, repayment and RFM (Recency, Frequency and Monetary) variables were defined. For classifying the legal customers of Export Development Bank of Iran in terms of credit, 5 variables were extracted from the bank’s database and normalized accordingly. The weight of each variable was calculated through interviewing bank experts using FAHP. Using the values of the variables and K-means algorithm, the optimal clusters of customers were determined. Finally, bank customers were ranked in 5 credit clusters and the value of each cluster was estimated.

According to the findings, recency, repayment behavior, transaction, frequency, and monetary variables had maximum effects on customers’ ranks, respectively. Therefore, 54% of the customers fell in the third cluster (with cluster value of 0.95) and the fifth cluster (with cluster value of 0.76) composed of good and very good customers. Credit risk of the two clusters (especially the third one) was at least. 32% of the customers positioned in the second cluster (with cluster value of 0.59) including the average customers in terms of credit. 14% of the customers fell in the fourth and first clusters with cluster values of 0.42 and 0.26 including highest risky customers.