Volume 21, Number 1 Article by Manish Kumar and M Thenmozhi March, 2009
Forecasting Stock Index Movement using Support Vector Machines and Random Forest Method :
Forecasting financial market trends has been one of the most challenging tasks for researchers and practitioners. In the recent years, the use of machine learning methods for forecasting is more popular than using fundamental analysis, technical analysis and traditional time series models.Machine learning techniques can trace both linear and nonlinear patterns and the most popularly used machine learning approaches are Artificial Neural Network(ANN),Random Forest Method(RFM)and Support Vector Machines(SVM). Though ANN has been a very successful tool for time series modelling and forecasting, it may not converge to global solutions and may have the danger of the overfitting problem, apart from the difficulty in selecting a large number of controlling parameters.
Since SVM and RFM have been used to solve estimation problems and they exhibit excellent performance the authors apply SVM and RFM to predict the direction of the SandP CNX Nifty Index movement.They also examine whether these models have the ability to outperform ANN and traditional benchmark models like logit and discriminant analysis.The results of the study show that SVM outperformed RFM ANN and other traditional models used in this study and RFM is better than ANN discriminant and logit models.SVM emerged as a suitable model to learn the pattern of the SandP CNX Nifty Index of India.Further RFM is a feasible method for financial time series application and this inference can be verified in other markets as well.Although SVM was slightly better than RFM the two models can further be evaluated for different financial time series.Policy makers can explore the use of SVM and RFM in predicting stock index movement and examine the impact on other economic indicators of the country.
Reprint No 09104
