The Study of SVM Optimized by Culture Genetic Algorithm on Predicting Financial Distress

In the analysis of predicting financial distress based on support vector machine (SVM), the two parameters of SVM, c and sigma, which its value have important effect on the predicting accuracy, must be predetermined carefully. In order to solve this problem, this paper proposed a new culture genetic algorithm (CGA) to optimize the parameters of SVM. Through embedding GA into the cultural algorithm framework, this CGA algorithm constructed the population space and the knowledge space based on genetic algorithm. The two spaces evolved independently, at the same time, the population space continuously transferred the evolving knowledge to the knowledge space, and then the knowledge space to achieve global optimization. Additionally, the proposed CGA-SVM model that can automated to determine the optimal values of SVM parameters was test on the prediction of financial distress of listed companies in China. Then we compared the accuracies of CGA-SVM with other models (Standard SVM, GA-SVM and GA-BPN). Experimental results showed that the hybrid of CGA with traditional SVM can serve as a promising alternative for predicting financial distress.