A multi-layer perceptron model of credit scoring.

Abstract

In this study, a Multi-Layer Perceptron (MLP), one of several neural network architectures, is used to construct a credit-scoring model for the prediction of credit card applicants' credit-worthiness. There are three types of applicant characteristics that affect credit-worthiness. They are character, capacity and capital. Each characteristic has its input variables and the ones selected for character are Education, Sex, Age and Marital Status. The input variables selected for capacity are Total Income, Occupation, Years in Service, Number of Dependents and Number of Cards Held. Those for capital are Home Ownership and Number of Bank Accounts.