一种对无统计环境测量模型缺省值进行预测的新方法  

A Novel Prediction Model for the Missing Data of Environmental Measurement

作　　者：李梦龙[1] 王智猛[1] 马宁[1] 王晃[1] 郁凌庄[1] 

机构地区：[1]四川大学化学学院,成都610064

出　　处：《四川大学学报（自然科学版）》2003年第4期736-739,共4页Journal of Sichuan University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(29877016)

摘　　要：采用模式识别中的K最近邻法(KNN方法)对不满足于统计模型的数据阵中的缺省值进行预测,预测以样本值作为预测参照 由于原始数据阵没有训练集,故采用对原始数据进行最大似然主成分分析(MPCA),获得的结论与KNN方法处理后的数据的主成分分析结论相比较,结果表明,两套数据分析得出的主因子数、因子负载阵基本一致,而因子得分阵有细微的差别。The Knearest neighbor method (KNN) is proposed to predict the missing data of environmental measurement among the matrix not appearing normal distribution. The prediction is based on the sample observations. For there are lack of training sets, in order to assure the method' rightness, the results of the maximum likelihood principal component analysis (MPCA) to the original data and the principal component analysis to the worked data are compared. For two results which have the same number of the principal components, the same loading matrixes, and the score matrixes are much at one, we conclude the KNN method is suitable, and the result of the worked data can provide more information.

关 键 词：KNN最邻近法 最大似然主成分分析 缺省值预测 环境数据 

分 类 号：O651[理学—分析化学]