The Hybrid Credit Scoring Strategies Based on KNN Classifier

Abstract

The development of credit scoring model has been regarded as a critical topic. This study proposed four approaches combining with the KNN (K-nearest neighbor) classifier for features selection that retains sufficient information for classification purpose. Two UCI data sets and different models combined with KNN classifier were constructed by selecting features. KNN classifier combines with conventional statistical LDA, Decision tree, Rough set and F-score approaches as features preprocessing step to optimize feature space by removing both irrelevant and redundant features. The procedure of the proposed algorithm is described first and then evaluated by their performances. The results are compared in combination with KNN classifier and nonparametric Wilcoxon signed rank test will be held to show if there has any significant difference between these approaches. Our results suggest that hybrid credit scoring models are robust and effective in finding optimal subsets and the compound procedure is a promising method to the fields of data mining.