关键词: 主成分分析, BP神经网络, 遗传算法, 遗传神经网络, 股票指数预测

Abstract: Data forecast occupies an important position in the financial investment field，the selected input variables affect the speed and accuracy of forecasts，traditional methods of selecting input variables subjective，and forecast ineffective.Combine Genetic Algorithms（GA） with BP neural network，using GA’s global to search optimized BP network structure parameters，overcome the local convergence and other issues of BP algorithm effectively.Using Principal Component Analysis（PCA） to select input variables，and GA-BP hybrid modeling is applied to the Shanghai stock index prediction.Experimental results show that this method improves the prediction accuracy and achieves a better prediction.

Key words: Principal Component Analysis（PCA）, BP neural network, genetic algorithms, GA－BP neural network, stock forecasting