13 填充算法.docVIP

稀疏编码的最近邻填充算法* 苏毅娟1 , 程德波2 , 宗 鸣2 , 李凌2 ,朱永华3 广西师范学院，南宁530023；2.广西师范大学计算机科学与信息工程学院，广西桂林 541004 广西大学 计算机与电子信息学院, 南宁530004K-Nearest Neighbor Imputation，简称KNNI）的参数K值固定问题了研究，发现对缺失值填充时，K值固定很大程度上影响填充效果这一重构过程中考虑了样本之间的相关性能分析指标RMSE和相关系数的实验比较结果表明，该算法比KNNI算法的效果要好。该算法能很好的避免KNNI算法的缺陷，适用于预处理缺失值填充应用。缺失值填充RMSE；相关系数 中图分类号：TP181 K Nearest Neighbor Imputation based on Sparse Coding SU Yi-juan1,CHENG De-bo2, ZONG Ming2, LI Ling2 , ZHU Yong-hua3 (1.Guangxi Teachers Education University, Nanning, 530023, Guangxi, China; 2. Guangxi Normal University, Guilin, 541004, Guangxi, China; 3.School of Computer, Electronics and Information, Guangxi University, Nanning, 530004, China) Abstract: Aimed at the parameter K fixed issues of K-Nearest Neighbor Imputation (KNNI) algorithm to research, and found that when imputed the missing values, the fixed value of the parameter K results in a large extent influence of the imputation effect. Therefore, the K Nearest Neighbor based on Sparse Coding (KNNI-SC) algorithm is proposed to solve this problem. This proposed method reconstructed each missing sample with the training samples, in the reconstruction process to fully consider the correlation between samples; and used anorm to learn to ensure each missing samples is imputed by different number of training samples, so solve the parameter K value selection problem of KNNI algorithm. Performance comparison based on the data analysis of the experimental results indicators RMSE and correlation coefficients show that the algorithm is better than KNNI algorithm. The algorithm can well avoid the defects of KNNI algorithm, it is available to data preprocessing step that needs missing values imputation’s applications. Key words: Missing value Imputation; sparse coding; reconstruct; RMSE; correlation coefficient; data preprocessing 0 引言 在数据挖掘和机器学习应用中，经常因一些原因使数据不能获取时完整无缺，造成数据缺失。比如有些信息暂时无法获取，数据被遗漏、不能正常收集的信息、获取某些信息的代价太大等都可能导致数据缺失[1,2]。而且在一些工业领域缺失值的比率高达80%以上[3,4]。数据缺失会影响到从中抽取规则的正