Enhancing Heart Disease Diagnosis with Machine Learning Algorithms: An Evaluation on Imbalanced Data and Performance Metrics

The precise identification of heart disease is of the greatest significance in enhancing patient outcomes and reducing mortality rates. This research aims to evaluate the effectiveness of machine learning algorithms in the early detection of heart disease. This study employs the dataset of the BRFSS obtained from the Kaggle data archive. The dataset shows a significant imbalance in class distribution. The SMOTE technique was employed to deal with this issue. Hence, the EDA was performed in order to gain a comprehensive understanding of the dataset. Ten algorithms have been employed, utilizing hyper-parameter tuning and 5-fold cross-validation techniques in order to optimize their performance. The evaluation of the machine learning models were performed using five performance metrics. Xgboost appeared as the most effective model during the simulation, achieving an accuracy rate of 92.29%. Yet, both Logistic Regression and Artificial Neural Network shows comparable and satisfactory performance, achieving accuracy of 92.10% and 92.05%, respectively. Therefore, this research has the potential to significantly aid healthcare professionals in effectively treating heart disease, which will encourage the establishment of an efficient e-healthcare system in the near future.