An Enhanced Fuzzy K-means Clustering with Application to Missing Data Imputation

Abstract

In this paper an adjustment on the Fuzzy K-means (FKM) clustering method was suggested to improve the process of clustering. Also, a novel technique for missing data imputation was proposed and it was implemented twice: (1) using FKM and (2) using the Enhanced Fuzzy K-means (EFKM) clustering. The suggested model for imputing missing data consists of three phases: (1) Input Vectors Partitioning, (2) Enhanced Fuzzy Clustering, and(3) Missing Data Imputation. The implementation and experiments showed a clear improvement in the imputation accuracy in favor of the EFKM according to the value of RMSE.