Hybrid Techniques of Combinatorial Optimization with Application to Retail Credit Risk Assessment

Abstract

Hybrid techniques of combinatorial optimization are a growing research area, designed to solve complex optimization problems. In the first part of this paper, we focus on the methodological background of hybrid techniques of combinatorial optimization, paying special attention to the important concepts in the field of combinatorial optimization and computational complexity theory, as well as to hybridization strategies that are important in the development of hybrid techniques of combinatorial optimization. According to the presented relations among the techniques of combinatorial optimization, the strategies of combining them and the concepts for solving combinatorial optimization problems, this paper presents an example of the hybrid technique for feature selection and classification in credit risk assessment. This study emphasizes the importance of hybridization as a concept of cooperation among metaheuristics and other optimization techniques. The importance of such cooperation is confirmed by the results that are presented in the experimental part of the paper, which were obtained on a German credit dataset using the hybrid technique of combinatorial optimization based on a low-level relay strategy. The experimental results show that the proposed method outperforms, on the same dataset, the methods presented in the literature in terms of the average prediction accuracy.

Credit Scoring with AHP and Fuzzy Comprehensive Evaluation Based on Behavioural Data from Weibo Platform

“…Plenty of previous works discussed the credit scoring method. Credit scoring methods can be generalized into two categories: statistical techniques and artificial intelligence methods [1,17,18]. Artificial Neural Networks (ANNs) is becoming more popular in the field of credit scoring because of its related memory characteristics and generalization ability [19,20].…”