A Methodology for Calculating Customer Credit Score Based on Customer Lifetime Value Model

Abstract

Proper customer relationship management is among the facets that contribute to productivity at institutions. It is a requirement for customer relationship managers, especially at financial and credit institutions and at banks, to calculate and determine the customer’s creditworthiness and credit score. The aim of this study is to present a solution for calculating the customers’ value and their credit score without incurring the costs for collecting extra information. The primary source of data for this study is operation system database. Due to differences among operation systems, a comprehensive schema of the database is defined first. Only conventional indices and variables have been used in this schema, so that the presented solution can be generalized and will be applicable to most economic institutions. The calculation of the customer’s creditworthiness is performed with regard to the three variables of the “recency” of contact, the “frequency” of transactions, and the “monetary” amount. The collected data is divided into the two populations of “good” and “bad” customers. Variables from those two populations that possess significant differences are identified using statistical methods. Those variables are used in determining the customer’s credit score. Next, a solution is presented for comparing the efficiency of the models for the identification of the customer’s credit score. We will test and compare two statistical methods, the Logistic Regression model and the Fisher Discriminant Analysis, and two soft computing methods, the Multilayer Perception Network and the Vector Machine for determining the customer’s credit score. Additionally, a solution is offered for setting the number of layers and the number of neurons in the Multilayer Perception Network.