Different forecasting models can complement each other in capturing patterns of data sets, and both theatrical and empirical studies have concluded that a combination of forecast outperforms individual forecasting models [14–16]. Since the early work of Bates and Granger [17], several architectures of combined forecasts have been explored. Clemen [18] had a comprehensive bibliography review in this area. Menezes et al. [19] offered good guidelines for combined forecasting. They concluded that the problem of combined forecasts is implementing multi-criteria process and judging the attributes of an error specification. Lam et al. [20] proposed a goal programming model to obtain optimal weights for combining forecasting models. Terui and Dijk [21] presented a linear and nonlinear time series model for forecasting the US monthly employment rate and production indices. Their results demonstrated that the combined forecasts outperformed the individual forecasts. Fang [22] used quarterly UK consumption expenditure data to show the superiority of the combined forecasting model. Zhang [23] combined the ARIMA and feedforward neural networks models in fore
casting. This study presents a hybrid model of ARIMA and SVMs to slove the stock price forecasting problem.