Ischemic Heart Disease Multiple Imputation Technique Using Machine Learning Algorithm

Medical datasets in profound data repository like the University of California Irvine (UCI) have missing values. These essential data are used for multiple analyses by researchers in a variety of predictions even though the results could be biased at times. It necessitates an important practice to use missing data imputation methods to fill up missing values for arriving at validated experimental results ensuring unbiased outcomes and predictions especially when the heart disease data set is handled. These methods are a type of treatment for data sets that include uncertainty and vagueness. Methods based on fuzzy-rough sets, on the other hand, offer excellent tools for dealing with ambiguity, with desirable properties such as robustness and noise tolerance. Fuzzy sets can also find minimal data representations and do not need potentially erroneous user inputs which confirms using fuzzy-rough sets for imputation be viable. In this paper we propose a novel Ischemic Heart Disease Multiple Imputation Technique (IHDMIT) missing value imputation methods based on fuzzy-rough sets and their recent extensions. The proposed IHDMIT with Random Forest classifier is compared with fuzzy roughest, fuzzy C means, and expectation maximization. The result shows that the proposed IHDMIT random forest classifier gives better accuracy of 93%.