Search

With an approximate 17.9 million annual victims, heart disease stands out as a prominent cause of deaths worldwide, whose fatality can be reduced to a great extent with an expeditious diagnosis. Herein we propose an ensemble learning-based heart disease prediction system. The UCI Heart Disease dataset has been utilized in this work. Relevant data mining methodology has been adopted to create six predictive models. Appropriate hyperparameters were optimized with the help of GridSearchCV along with 5-fold cross-validation. Recall value and ROC score were the performance metrics considered relevant to judge the performance of the models. The best performing models were picked to create a heterogeneous ensemble. The proposed ensemble produced a ROC score of 0.84 and a recall value of 0.94. The suggested ensemble has been found to enhance the predictive capabilities of the classic algorithms