摘要
在误差逆传播算法神经网络预测模型数据前处理中,对样本集优化,采用多元统计分析中的主成分分析法,提取影响粉煤灰高性能混凝土抗压强度的主要因素,消除影响因素间的线性相关性。研究结果表明,用该方法处理后的样本数据输入神经网络,提高了预测效率,训练时间减少,预测精度也有一定程度的提高,网络结构得到简化。
The principal component analysis in multivariate statistic analysis is used to optimize the sample set in data preprocessing of prediction models of error back propagation neural networks.This method can extract main factors that effect on the compressive strength of high performance concrete containing fly ash and can eliminate the linear correla-tion among the factors.The result of study indicates that while the processed data by this method input the neural net-works,the efficiency of forecast is improved,the training time is reduced,the prediction accuracy is also developed,and the structure of neural networks becomes simple.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第18期192-195,共4页
Computer Engineering and Applications
基金
国家自然科学基金资助项目(编号:50178028)
关键词
混凝土抗压强度
多元统计分析
主成分分析
人工神经网络
compressive strength of high performance concrete,multivariate statistic analysis,Principal Component Analy-sis,neural networks