BP神经网络应用中的前后处理过程研究

在 BP网络的应用过程中 ,采用数理统计中的一些技巧 ,用主成分分析法作为前处理过程进行样本集的选择与优化 ,用回归分析和相关性分析的方法作为后处理过程之一对训练结果进行分析 ,用置信区间分析的方法作为另一种后处理过程对输出结果的可信度进行预测 .这几种处理过程可以作为 BP神经网络应用研究的一个必要的补充.

Research on Preprocessing and Postprocessing of the Application of BP Neural Network

During the application of BP neural network, some techniques in statistics are adopted. The principal component analysis is used as the preprocessing to select and optimize the sample set. The regression and correlation analysis is used as one of the postprocessing to analyze the result of training; the confidence interval analysis is used as the other postprocessing to predict the confidence bounds of output. All of these processes can be a necessary supplement of the research on the application of BP neural network.