Three local search-based methods for feature selection in credit scoring

Abstract

Credit scoring is a crucial problem in both finance and banking. In this paper, we tackle credit scoring as a classification problem where three local search-based methods are studied for feature selection. The feature selection is an interesting technique that can be launched before the data classification task. It permits to keep only the relevant variables and eliminate the redundant ones which enhances the classification accuracy. We study the local search method (LS), the stochastic local search method (SLS) and the variable neighborhood search method (VNS) for feature selection. Then, we combine these methods with the support vector machine (SVM) classifier to find the best described model from a dataset with the correct class variable. The proposed methods (LS+SVM, SLS+SVM and VNS+SVM) are evaluated on both German and Australian credit datasets and compared with some well-known classifiers. The numerical results are promising and show a good performance in favor of our methods.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 31 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.