Comparison of Clustering Performances of Missing Data Imputation Methods

Abstract

In observational or measurement-based research, if the data cannot be accessed for any reason or the accessed data cannot be recorded, a missing data problem occurs. In the analysis of this problem, missing data generation mechanisms are used. According to the relationship between the probability of missing data and the data set, the missing data generation mechanisms have three basic structures. By determining the missing data mechanism, analyzing the studies containing missing data gives more accurate results. After determining the missing data mechanism, two different methods for missing data; It is either deleting the rows of the missing values from the data set or imputing the missing values. In this study, according to the World Health Organization (WHO); Stroke data set, which is the second leading cause of death, is responsible for approximately 11% of the total mortality rates and is a brain disorder, has been studied. Missing Completely Random - Artificial missing data sets produced at 7 different rates over the full data set with the MCAR Data Mechanism were completed with 6 different imputation methods, and clustering prediction performances based on the full data set were comnared. © 2021 IEEE.