Performance Analysis of Imputation Approaches with Noise Filtration on Medical Data

Accuracy of any prediction model on the data set which is having large number of missing values is always a challenging problem. The popular technique in order to handle missing values is known as case deletion in which we simply delete the missing instances. Many hybrid models have either used some default way to fill the missing values or case deletion. Concept Most Common (CMC) method has proved best imputation technique in case of Hepatitis data set and Case deletion has performed best in case of Wisconsin Breast Cancer Data, according to literature. The effectiveness of this missing value imputation is heavily dependent on the observed data (or complete data) and that data too contains many imperfections like noise, redundancies etc. So, the objective of this paper is to investigate the effect of performing noise filtration on the imputation task. Specifically, two different procedures for combining instance selection and missing value imputation are proposed and are compared on the basis of their classification accuracy. The results obtained are promising and superior to the recently proposed methods.