Development of Discriminant Analysis and Majority- Voting Based Credit Risk Assessment Classifier

摘要

This article presents a research on a method for credit risk evaluation combining expert majority-based ensemble voting scheme together with discriminant analysis as basis for expert formation and popular machine learning techniques for classification, such as decision trees, rule-based inducers and neural networks. Both single expert and multiple expert evaluations were applied as basis for forming output classes dynamically. Feature selection was applied using correlation-based feature subset evaluator with tabu search. The experiment results form a basis for further research of similar method.