Forecasting stock index trend based on the GAS-VM integrated system and wavelet-based feature extractions on multiple scales

This paper proposes a novel GA-SVM integrated system for stock trend prediction based on wavelet-based feature extractions on multiple scales. The parameters of support vector machine (SVM) and kernel function are optimized by Genetic Algorithm (GA). Wavelet transformation is used to form the wavelet-scaling features. The Shanghai Stock Exchange (SSE) Composite index is selected for this study. Sufficient experiments are carried out, resulting in significant performances of the novel GA-SVM integrated system based on the wavelet-based feature selection method.