A Credit Scoring Model Based on Collaborative Filtering

Abstract

To ensure property safety, risk assessment plays an essential role in modern society. Credit scoring, which is a significant branch of exposure rating, becomes a hot topic. As a result, various kinds of credit scoring models are established to evaluate the customers' credit rank. In this paper, a simple credit scoring model, Collaborative Filtering based on Matrix Factorization with data whose continuous attributes are discretized considering Information Entropy (CF-MF-D-IE), is constructed to solve credit scoring issues. The proposed model is tested on two important credit data sets in UCI Repository of Machine Learning databases. Compared with Collaborative Filtering using non-discretized data and Support Vector Machines with discretized data, CF-MF-D-IE has better classification accuracy rate.