基于因子分析的BP神经网络在需水预测中的应用

Application of BP neural network based on factor analysis in prediction of water requirement

需水预测涉及农业、工业、生活和环境等多种因素,是典型的多指标评价问题,需对多因素进行综合评估.以广东省珠海市为例,利用1986～2000年的需水量数据,采用因子分析法对影响需水量的8个变量进行因子分析,根据确定的主要影响因子构造BP神经网络的输入样本,从而进行不同水平的需水量预测.结果表明:前4个公因子,即综合实力因子、畜牧影响因子、环境影响因子和补充分析因子,累计贡献率达99%以上,为影响研究区需水量的主要因子,以此作为输入样本建立的BP神经网络需水预测模型既能合理地构造神经网络的拓扑结构,又可加快网络的收敛速度,精度较高.因子分析与BP神经网络结合是多因素需水预测的一个有益尝试.

Prediction of water requirement relates to agriculture,industry,life,environment and other factors.It is a typical problem involved in multiple evaluations,which need to be integrated evaluated by a number of factors.Taking Zhuhai City,Guangdong Province as an example,8 variable factors which influence water requirement were used as analysis factors.The base data were from 1986 to 2000.According to input samples of the main influencing factor structure of BP neural network,we can carry out water demand predictions in different levels.Results indicate that the first 4 common factors which have the cumulative contributions of more than 99%,are the main factors which influence the water requirements of an area.So BP neural network based on this input sample could not only construct water demand prediction model of neural network topology reasonably,but also accelerate the convergence of network speed,which has higher accuracy.The combination of factor analysis and BP neural network is an useful attempt of water requiremnet prediction of multiple factors.