Forecasting the Survival and Mortality of Patients by Machine Learning Trained on Heart Failure Clinical Imbalanced Data

A heart attack arises when blood flow in a heart is strictly reduced or congested. The blockage generally comes from the accretion of cholesterol, sugar, fats, etc., and other substances in the arteries (coronary) of the heart. Machine learning is a fantastic tool for estimating the possibility of a heart attack or patient survival based on prior medical data. The search indicates that few research articles assess the impact of imbalanced data distribution on disease prediction. In our research, we use the Heart Failure Clinical Records dataset to train and test five well-known machine learning algorithms, such as Decision Tree, Random Forest, AdaBoost, CatBoost, and LightGBM, separately on two different scenarios: the original dataset and the artificially resampled dataset, which are then used to forecast the survival of patients or deaths. The Synthetic Minority Oversampling Technique with the Edited Nearest Neighbor method is used to balance the data distribution synthetically, and then the Random Forest model of feature selection is applied to select the optimal features of the dataset. Performance metrics, including recall, accuracy, precision, F1-score, G-mean, and AUC, are used to analyze findings. When the original dataset was used, results indicated that the Random Forest classifier outperformed other classifiers on all performance metrics and attained the highest AUC value of 0.86. In contrast, when using the resampled dataset, the LightGBM classifier outperformed other classifiers on all performance metrics and attained the highest AUC value of 0.960.