Feature selection for support vector machine in financial crisis prediction: a case study in China

Corporate financial crisis forecasting plays an increasingly important role in the current intense and competitive commercial environment. It is an important phase in financial crisis forecasting to find the features that discriminate different financial conditions. In this paper, a genetic algorithm (GA) based approach and statistical filter approaches are applied to identify the best features for the support vector machine (SVM). The proposed GA-based approach is carefully designed in order to have the capability of simultaneously optimizing the features and parameters of the SVM. Experimental results on the data from Chinese companies show that the GA-based approach can extract fewer features with a higher accuracy compared with statistical filter approaches, such as analysis of variance, the T-W test (which is the t test applied to variables satisfying a normal distribution and the Wilcoxon test applied to other variables not satisfying a normal distribution), logit regression and multiple discriminant analysis. Moreover, the experiments indicate that the proposed GA-based approach is robust and suitable for selecting features for the SVM.