Forecasting CSI 300 index using a Hybrid Functional Link Artificial Neural Network and Particle Swarm Optimization with Improved Wavelet Mutation

Financial market dynamics forecasting has long been a focus of economic research. A hybridizing functional link artificial neural network (FLANN) and improved particle warm optimization (PSO) based on wavelet mutation (WM), named as IWM-PSO-FLANN, for forecasting the CSI 300 index is proposed in this paper. In the training model, it expands a wider mutation range while apply wavelet theory to the PSO, in order to exploring the solution space more effectively for better parameter solution. In the stimulating experiment, we use five benchmark functions to test the proposed method, and the results shows that IWM-PSO has greater convergence accuracy than WM-PSO and PSO. The empirical research is performed in testing the predictive effects of CSI 300 index in the proposed model compared with the back propagation functional link neural network (BP-FLANN), PSO-FLANN and WM-PSO-FLANN. The experiment utilizes two expansion functions, Chebyshev functions and trigonometric functions, to map the input data to higher dimension. The results show that the prediction performance of the proposed model displays a better performance in financial time series forecasting than other three models. Moreover, the accuracy of the input with trigonometric functions is higher, and it suggests that trigonometric function is more suitable for this kind of data type.