Tehran Stock Price Modeling and Forecasting Using Support Vector Regression (SVR) and Its Comparison with the Classic Model ARIMA

Use of linear and non-linear models to predict the stock price has been paid attention to by investors, researchers and students of finance and investment companies, and organizations active in the field of stock. Timely forecasting stock price can help managers and investors to make better decisions. Nowadays, the use of non-linear methods in modeling and forecasting financial time series is quite common. In recent years, one of the new techniques of data mining with support vector regression (SVR) has had successful application in time series prediction. In this study, using support vector regression model, we examined the Tehran Stock prices and the predicted results were compared with ARIMA classic model. For this purpose, of the Tehran stock companies, 5 companies were selected during the years 2002 to 2012. Using benchmarks to evaluate the performance of MSE, MAE, NMSE these two methods were compared and the results (in the case of non-linear data) indicate the superiority of SVR model compared to the ARIMA model.