The Application of Modern Optimization Algorithm in Time Series Prediction
Zhaoyue Hu, Yanping Bai · 2015
This paper applies the SC model and SVJD model to artificially generated data, and we put forward a forecast model that hybridizes genetic algorithm, principal component analysis and artificial neural network methods.This article utilizing genetic algorithm to search for the initial weights of the BP neural network could guarantee a relatively high probability to obtain the global optima, and we include principle component analysis (PCA) to extract contribution rate to meet 85% of the principal component as the input variables.The experiment results demonstrate that the combination methods PCA-BP and PCA-GA-BP model is adopted to overcome the fitting compared with the traditional forecasting method.