Integrating GA-based time-scale feature extractions with SVMs for stock index forecasting

By integrating genetic algorithm (GA)-based optimal time-scale feature extractions with support vector machines (SVM), this study develops a novel hybrid prediction model that operates for multiple time-scale resolutions and utilizes a flexible nonparametric regressor to predict future evolutions of various stock indices. The time series of explanatory variables are decomposed using wavelet bases, and a GA is employed to extract optimal time-scale feature subsets from decomposed features. These extracted time-scale feature subsets then serve as an input for an SVM model that performs final forecasting. Compared with neural networks, pure SVMs or traditional GARCH models, the proposed model performs best. The root-mean-squared forecasting errors are significantly reduced.