Forecast of Stock Index Volatility Using Wavelet Support Vector Machines

Support vector machines (SVM) is one of the important intelligent forecasting methods. Wavelet support vector machines (WSVM) is a kind of SVM, which replace the ordinary kernel function in SVM as wavelet kernel function. Using range volatility as the proxy of the true volatility, this paper applied WSVM with three different wavelet kernel functions to forecasting Shanghai Composite Index (SHCI) in China stock markets. In-sample and out-of-sample forecasting results from the three WSVM models are compared with those of SVM model with Radial basic kernel function (RBF) to test the effectiveness of WSVM. Empirical results indicate that for the evaluation indices HRMSE and HMAE, WSVM with three wavelet kernel functions are superior to SVM with RBF kernel function for in-sample and out-of-sample volatility forecasting accuracy. And the WSVM using Morlet kernel function has better volatility forecasting performance than the other two WSVM models.