MACHINE LEARNING BASED HEART DISEASE PREDICTION SYSTEM

In today's world, heart disease remains one of the leading causes of death globally. The medical field continuously gathers vast amounts of data to diagnose diseases, but not all collected information is relevant for making accurate decisions. Detecting the early onset of heart disease can be challenging as it requires expert knowledge and experience in recognizing symptoms. Additionally, medical datasets are often scattered, diverse, and heterogeneous, making analysis complex. Data mining plays a crucial role in extracting hidden, meaningful, and predictive insights from large databases. This paper utilizes information gain-based feature selection to eliminate irrelevant data and applies various classification techniques—such as K- Nearest Neighbors (KNN), Decision Tree-ID3, Gaussian Naïve Bayes, Logistic Regression, and Random Forest—on a heart disease dataset to improve prediction accuracy. The classification models are evaluated using performance metrics, including accuracy, ROC curve, precision, recall, sensitivity, specificity, and F1-score. Among these techniques, the Decision Tree model demonstrated the highest accuracy, achieving approximately 97% classification accuracy.