An Improved Decision Tree Approach Combined with Hybrid Feature Selection Methods for Prediction of Heart Disease

Background: Heart disease diagnosis poses significant challenges in clinical settings, contributing to rising mortality rates. With the increasing availability of medical datasets and massive data, accurately predicting cardiac disease has become more complex due to extraneous features and computational challenges. Objective: This research aims to improve the accuracy of heart disease prediction by proposing a feature selection-based enhancement to decision tree classification methods. Methods: The study employs a hybrid feature selection approach to eliminate irrelevant features and address missing values in the dataset. An improved decision tree method, incorporating feature ranking, is proposed to optimize model inputs and reduce dimensionality. Data for the study were obtained from the UCI Machine Learning Repository, specifically the heart disease dataset. Performance evaluation utilized metrics such as accuracy, precision, recall, and Fscore, commonly applied in medical classification studies. Results: The proposed hybrid feature selection and improved decision tree method demonstrated superior efficiency compared to other classification algorithms. The method achieved an accuracy improvement of approximately 1.52% over existing approaches, highlighting its effectiveness. Conclusion: The integration of hybrid feature selection and an enhanced decision tree classifier offers a promising solution for accurate heart disease prediction, addressing key challenges associated with extraneous features and computational complexity in medical datasets.