Correlated Cluster-Based Imputation for Treatment of Missing Values

Improved imputation has a major role in the research of data pre-process for data analysis. The missing value treatment is implemented with many of the traditional approaches, such as attribute mean/mode, cluster-based mean/mode substitution. In these approaches, the major concentration is missing valued attribute. This paper presents a framework for correlated cluster-based imputation to improve the quality of data for data mining applications. We make use the correlation analysis on data set with respect to missing data attributes. Based on highly correlated attributes, the data set is divided into clusters using suitable clustering techniques and imputes the missing content with respect to cluster mean value. This correlated cluster-based imputation improves the quality of data. The imputed data are analyzed with K-Nearest Neighbor (KNN) and J48 Decision Tree multi-class classifiers. The efficiency of imputation is ascertaining 100 % accuracy with correlated cluster mean imputed data compared with attribute mean imputed data.