An Effective Hybrid GA-PP Strategy for Artificial Neural Network Ensemble and Its Application Stock Market Forecasting

摘要

The learning and generalizing ability of artificial neural network dependents on the particular training set. In this study, a novel hybrid GA-PP strategy for neural network ensemble model is proposed for stock market forecasting. First of all, we use the Projection Pursuit Technology based on Genetic Algorithms optimized to extract input factors, and then many individual neural networks are generated by Bagging techniques and different training way. Secondly, Projection Pursuit Technology based on Genetic Algorithm is used to select appropriate ensemble members. Finally, the logistic regress method is used for neural network ensemble. This method is established to forecast the Shanghai Stock Exchange index. The result shows that the ensemble network has reinforced the learning capacities and generalizing ability.