基于改进的主成分分析法对水质定量化综合评价

A Comprehensive Quantitative Evaluation of Water Quality Based on Improved Principal Components Analysis

为客观、准确地评价饮用水水质综合状况，针对传统主成分分析法对数据非线性处理效果不理想及综合评价时表达信息的片面性，提出相应的改进方法：对原始数据进行中心化对数比处理，采用协方差矩阵代替相关系数矩阵，充分反映数据的变异性特征．在保留原始管网信息的基础上得到表征水质状况的综合指标量。结合实例对比改进前后2种方法的综合评价效果，证明改进的主成分分析法更能真实反映水质情况。

To evaluate the comprehensive situation of drinking water quality objectively and accurately, a corresponding improved method was proposed in view of the unsatisfactory processing result in dealing with nonlinear data and the one-sidedness when expressing information in the traditional principal components analysis. The original data were processed with centralised logarithmic ratio method and the covariance matrix was used instead of correlation coefficient matrix so as to reflect the variability of data adequately. A composite indicator characterising water quality has been obtained whilst the original pipe network information was maintained. With a concrete comparison of comprehensive evaluation effects between the methods before and after the improvement, it has been shown that the improved principal components analysis was better to reflect the real situation of water quality.