The Use of Genetic Algorithm, Clustering and Feature Selection Techniques in Construction of Decision Tree Models for Credit Scoring

Mohammad Khanbabaei,Mahmood Alborzi

doi:10.5121/ijmit.2013.5402

Abstract

The Use of Genetic Algorithm, Clustering and Feature Selection Techniques in Construction of Decision Tree Models for Credit Scoring

Highlights

Gary and Fan (2008) [1] believed that, "Banks as economic institutions need to recognize customers’ credit risk to offer credit facilities and manage their risk"
The main problem is the construction of decision trees to be able to classify bank customers optimally
This study has proposed a new hybrid classification model for designing a customer credit scoring model for banks

Summary

Introduction

Gary and Fan (2008) [1] believed that, "Banks as economic institutions need to recognize customers’ credit risk to offer credit facilities and manage their risk". Non-parametric methods and data mining have been used in the customers’ credit scoring techniques. As one of the classification techniques in data mining, can help to perform customer credit scoring with high ability of understanding and learning speed to build classification models. The main problem in this study is the construction of decision trees to classify bank customers optimally. Greediness in the tree growing process and local optimization at each step in the node splitting. 2. Tendency to construct large trees, and over-fit to training datasets [1], and generalization problem [2]. According to greedy rule induction algorithms, there is a single candidate solution every time and the evaluation is performed in a special candidate solution (based on local optimization). Using probabilistic operators, GA prevents solutions to be locked in local optimization [2]

Objectives

Results

Conclusion