Abstract
The quality of data is determined by several things, namely the completeness and balance data. The heart disease dataset from the University of California, Irvine (UCI) has missing and imbalanced data, which if it is not handled, can lead to a lack of accuracy in the prediction model and errors in interpreting the data. To overcome missing data, several methods can be used, one of which is data imputation. Attributes with missing data of 5% or less are handled using imputation methods such as Mean, Mode, and MICE. Attributes with numeric types are handled by Mean. Attributes with categorical types are imputed byMode. Attributes with more than 5% missing data are imputed using the MICE method. Imbalanced data can be handled by applying an oversampling method using the Adaptive Synthetic Sampling Approach (ADASYN). The effect of imputing missing data and addressing class imbalance on heart disease classification performance was tested using Random Forest, Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP) algorithms. After handling missing values and data imbalance, improvements were observed in the classification results. The accuracy, precision, recall, and F1-score showed excellent performance, above 90% on several classification methods. The results indicate that handling missing and imbalanced data through Mean, Mode, MICE, and ADASYN positively impacts the performance of classifiers on the UCI heart disease dataset.