Abstract: Based on the panel data of China’s capital market， the financial fraud companies from 2006 to 2015， together with the same number of non-fraud companies were selected as the research samples. Twenty-seven financial and non-financial indexes were analyzed， after which the dimension of the indexes was reduced through the test of independence and eight indexes were retained as the modeling parameters. The grid search algorithm， genetic algorithm and particle swarm optimization(PSO)were used respectively to optimize the parameters， and three support vector machine(SVM)models with the parameters optimized by the proposed methods were established respectively for financial fraud detection. The results showed that the SVM model with the parameters optimized by PSO has a higher detection rate than the other two models.

Key words: parameter optimization, support vector machine(SVM), financial fraud, detection model, genetic algorithm(GA), particle swarm optimization(PSO)