Using Support Vector Machines in Financial Time Series Forecasting

Forecasting financial time series, such as stock price indices, is a complex process. This is because financial time series are usually quite noisy and involve ambiguous seasonal effects due to holidays, weekends, irregular closure periods of the stock market, changes in interest rates, and announcements of macroeconomic and political events. Support vector machines (SVM) and Artificial neural networks (ANN) have been used in a variety of applications, mainly in classification, regression, and forecasting problems. In the SVM method for both regression and classification, data is mapped to a higher-dimensional space and separated using a maximum-margin hyperplane. This paper investigated the application of SVM in financial forecasting. The autoregressive integrated moving average (ARIMA), ANN, and SVM models were fitted to Al-Quds Index of the Palestinian Stock Exchange Market time series data and two-month future points were forecast. The results of applying SVM methods and the accuracy of forecasting were assessed and compared to those of the ARIMA and ANN methods through the minimum root-mean-square error of the natural logarithms of the data. We concluded that the results from SVM provide a more accurate model and a more efficient forecasting technique for such financial data than both the ANN and ARIMA models.