A machine learning approach combining expert knowledge with genetic algorithms in feature selection for credit risk assessment

Abstract

Most credit scoring algorithms are designed with the assumption to be executed in an environment characterized by an automatic processing of credit applications, without considering the input of expert opinions. Since in credit scoring applications expert opinions have been proved very helpful, in this work, we propose a combination strategy of integrating soft computing methods with expert knowledge. The ability of interpretation of the predictive power of each feature in the credit dataset is strengthened by the engagement of experts in the credit scoring process, and the proposed wrapper-based feature selection approach which explores how the features contributing most towards the classification of borrowers. In particular, an unsupervised machine learning algorithm allows experts to create one or more clustering scenarios regarding the features by defining a desired number of clusters per scenario. For each clustering scenario, the Analytic Hierarchy Process is applied for helping experts to make preferences for features of each cluster, set pair-wise constraints for features of equal importance, and evaluate their subjective judgments in terms of consistency. Then, expert opinions are taken into consideration for solving a credit scoring problem in the form of an optimization problem subject to constraints via soft computing methods based on supervised machine learning and evolutionary optimization algorithms. Testing instances on standard credit dataset are established to verify the effectiveness of the proposed methodology.