Forecasting Range Volatility Using Support Vector Machines with Improved PSO Algorithms

Abstract

Listen

Translate article

Financial volatility forecasting is an important content in financial derivatives pricing, financial risk management, portfolio allocation. Support vector machines (SVM) has been successfully used in modeling and forecasting volatility. To enhance the volatility forecasting ability of SVM, this paper introduced three improved particle swarm optimization (IPSO) algorithms into SVM for forecasting range volatility. SVM was used to construct the volatility forecasting model. And, the three IPSO algorithms: stochastic inertia weight PSO (SIWPSO), time-varying acceleration coefficients PSO (TVACPSO) and two-order oscillating particle swarm optimization (TOOPSO) were applied to finding the optimal parameters in SVM, respectively. The accuracy of the proposed models in forecasting range volatility was demonstrated by using three stock indices in China stock market. The empirical results imply that compared to SVM with SIWPSO and TVACPSO algorithms, SVM with TOOPSO algorithm produces higher accuracy in forecasting range volatility and faster speed on searching for the optimal parameters of SVM.