Rank Aggregation for Filter Feature Selection in Credit Scoring

Abstract

Abstract. The credit industry is a fast growing field, credit institutions collect data about credit customer and use them to build credit model. The collected information may be full of unwanted and redundant features which may speed down the learning process, so, effective feature selection methods are needed for credit dataset. In general, Filter feature selection methods outperform other feature selection techniques because they are effective and computationally fast. Choosing the appropriate filtering method from the wide variety of classical filtering methods proposed in the literature is a crucial issue in machine learning. So, we propose a feature selection fusion model that fuses the results obtained by different filter feature selection methods via aggregation techniques. Evaluations on four credit datasets show that the fusion model achieves good results.

A feature selection model based on genetic rank aggregation for text sentiment classification

“…Bouaguel et al [20] presented a rank aggregation-based feature selection method for credit scoring domain. Relief, Pearson correlation coefficient and mutual information methods have been used as the individual feature selection methods.…”

Comparison of rank aggregation methods based on inherent ability

“…This problem has received significant attention recently (Köksalan, ; Lansdowne, ). Rank aggregation has penetrated many areas of decision‐making and evaluation, such as meta‐search engines (Dwork et al, ), voting systems (Obata & Ishii, ), and credit scoring (Bouaguel, Mufti, & Limam, ). Therefore, information providers and managers who rely heavily on new technology are paying significant attention to developing effective rank aggregation methods to identify the best alternatives.…”