Predictive Modelling of Cardiac Disease: Enhancing Accuracy Through Machine Learning Algorithms and Borderline-SMOTE Technique

In the field of medicine, the early detection of cardiac issues presents a considerable hurdle. Globally, this disease consistently stands as one of the leading causes of mortality. The prompt identification and precise prediction of heart disease are crucial for effective disease management and prevention. Nevertheless, conventional approaches have struggled to enhance the performance of classifying cardiac diseases. Machine learning (ML) and deep learning (DL) are widely recognized methods to enhance the performance of classifying cardiac diseases. This research work uses a kaggle heart disease dataset to predict the cardiac disease via various ML algorithms such as logistic regression (LR), Gaussian Naïve Bayes (GNB), Random Forest (RF), Support Vector Machine (SVM), K-nearest Neighbour (KNN), Ada Boost, CatBoost, Gradient Boosting (GB) algorithms. Furthermore, the given dataset is imbalanced, which has the potential to diminish overall performance. The proposed work uses a Borderline_SMOTE (BS) to handle given imbalance data. These results revealed that GNB model gives highest accuracy of 80.98%, precision of 87.50, Recall of 78.50%, F1 Score of 82.76% and Specificity of 84.42%. GNB gives highest performance results compare to other algorithms. CatBoost gives highest ROC-AUC curve value (0.93). These results indicate that the BS-based GNB algorithm demonstrated superior performance compared to other ML models.