Combination of feature selection approaches with random binary search and Mahalanobis Taguchi System in credit scoring

Abstract

The wide usage of credit scoring is to determine credit, credit analysis cost reduction, and faster decision making. However, traditional credit scoring models are not the factor for the influence of noises. There are several objectives of this study: 1) to develop Random Binary Search algorithm-based feature selection in Mahalanobis Taguchi System (MTS), 2) to validate the techniques of feature selection problems which are computationally efficient, and 3) to apply Random Binary Search algorithm in solving credit problems. In this study, Random Binary Search (RBS) algorithm was proposed, which was incorporated between MTS. The purpose of this was to optimize the procedures for determining the most useful variables. Besides being a relatively new statistical methodology where various mathematical concepts are integrated, there is an involvement of MTS in the field of diagnosis and classification in multidimensional systems. Besides, it is a highly effective method and has been implemented to a range of disciplines such as engineering, medical, financial, and others. As for the main problem of this study, data sets of credit scoring were used.