An Improved Heart Disease Prediction Using Information Gain-Based Feature Selection

Heart disease remains one of the leading causes of mortality worldwide, accounting for a significant proportion of deaths annually. Early and accurate prediction of heart disease risk is therefore essential for guiding timely clinical intervention and reducing healthcare burdens. However, predictive models often suffer from reduced performance due to redundant and irrelevant features present in medical datasets. This study addresses this challenge by applying Information Gain-based feature selection to improve the reliability of heart disease prediction. The research utilized the Kaggle Heart Disease Dataset, which consists of demographic and clinical attributes including age, sex, chest pain type, resting blood pressure, cholesterol level, exercise-induced angina, and ST-slope characteristics. Information Gain, an entropy-based ranking criterion, was employed to identify and retain the most informative features while discarding less relevant variables. By reducing dimensionality, the approach enhanced both model interpretability and computational efficiency. Experimental evaluation demonstrated that models trained on the Information Gain-selected features achieved higher accuracy and better generalization compared to models trained on the full dataset. The feature selection process also highlighted the clinical risk factors most strongly associated with heart disease, such as chest pain type, ST-slope, and exercise-induced angina. The results confirm that Information Gain-based feature selection significantly improves predictive performance and provides valuable insights into the attributes most indicative of heart disease risk. This approach contributes to the development of lightweight, interpretable, and effective predictive systems that can support clinical decision-making and early diagnosis.