Research on Forecasting Approach for Complex Time Series Based on Support Vector Machines

The technology of phase space construction and Support Vector Machines(SVM) is introduced firstly. Then a novel complex time series forecasting approach based on SVM is proposed. The complex time series is decomposed into long-term trend series and short-term fluctuation series. The SVM regressive forecasting model is constructed respectively. The proposed forecasting approach is applied to the Shanghai stock index data and the parameter sensitivity of SVM is analyzed. Experimental results indicate that the proposed forecasting approach is effective for complex time series.