Stock trend prediction based on fractal feature selection and support vector machine

Stock trend prediction is regarded as a challenging task. Recently many researches have shown that a successful feature selection method can improve the prediction accuracy of stock market. This paper hybridizes fractal feature selection method and support vector machine to predict the direction of daily stock price index. Fractal feature selection method is suitable for solving the nonlinear problem and it can exactly spot how many important features we should choose. To evaluate the prediction accuracy of this method, this paper compares its performance with other five commonly used feature selection methods. The results show fractal feature selection method selects the relatively smaller number of features and it achieves the best average prediction accuracy.