Abstract – According to a recent WHO report, heart-related problems are growing more widespread, with 17.9 million people suffering as a result each year. With a growing demographic, it becomes more difficult to detect and initiate treatment at an early stage. This arouses great interest in professionals in the field and researchers in creating more effective and reliable predictors. This study presents an automatic and proficient approach for predicting heart disease by introducing a feature selection-based machine learning technique. We integrated feature selection techniques, specifically the Correlation matrix and least absolute shrinkage and selection operator (LASSO). After Training all models, the Random Forest Algorithm achieved significant performance, with an accuracy rate of 97.48%. We also experimented with deep neural networks with an accuracy of 77.72%. The results were compared to previous studies, which indicated that performance was not significantly better than the model we provided. So, the provided technology accurately predicts heart disease, which will be employed by the healthcare field to assist people in reducing their death rates.
Cite: Arifuzzaman, M., Chowdhury, M. J. U., Ahmed, I., Siddiky, M. N. A., & Rashid, D. (2024, July). Heart Disease Prediction through Enhanced Machine Learning and Diverse Feature Selection Approaches. In 2024 IEEE 10th International Conference on Smart Instrumentation, Measurement and Applications (ICSIMA) (pp. 119-124). IEEE.