Credit risk evaluation modeling using evolutionary linear SVM classifiers and sliding window approach

This paper presents a study on credit risk evaluation modeling using linear Support Vector Machines (SVM) classifiers, combined with evolutionary parameter selection using Genetic Algorithms and Particle Swarm Optimization, and sliding window approach. Discriminant analysis was applied for evaluation of financial instances and dynamic formation of bankruptcy classes. The possibilities of feature selection application were also researched by applying correlation-based feature subset evaluator. The research demonstrates a possibility to develop and apply an intelligent classifier based on original discriminant analysis method evaluation and shows that it might perform bankruptcy identification better than original model.