基于LPP和l_(2,1)的KNN填充算法

KNN Imputation Algorithm Based on LPP and l_(2,1)

传统的KNN缺失值填充算法存在没有利用样本间属性的相关性,也没有考虑到保持样本数据本身的结构和去除噪声样本的问题。本文提出利用训练样本重构测试样本从而进行最近邻缺失值填充的方法,该方法重构过程充分利用样本间的相关性,也用到LPP(保局投影)保持数据结构在重构过程中不变,同时引入l2,1范式用于去除噪声样本。在UCI数据集上的仿真实验结果表明,该方法比传统的KNN填充算法以及基于属性信息熵的Entropy-KNN算法有更高的预测准确度。

Traditional KNN missing data filling algorithm does not utilize the correlation between the properties of samples,Neither considers but also does not consider to maintain the sample structures and removes noise samples.In this paper,a method of using training samples to reconstruct the test sample is proposed,which is used for the nearest neighbor missing data imputation.The method makes full use of the correlation between samples,uses the LPP（locality preserving projection）to maintain the data structure in the process of reconstruction,and uses l2,1norm to remove noise samples.Simulation experiments on UCI data sets show that the proposed method has higher prediction accuracy than the traditional KNN algorithm and Entropy-KNN algorithm based on attribute information entropy.