Hyperparameter Tuning in Machine Learning Classifiers: Performance Insights for CVD Prediction

Abstract

Globally, Cardiovascular Diseases (CVDs) continue to be a leading cause of morbidity and mortality, placing a significant burden on healthcare systems. Managing the risk factors linked to CVDs requires early identification and resolution. In this article, machine learning techniques (MLT) for predicting heart disease were presented. This approach presents the effects of dimension reduction and hyper-parameter adjustment on the prediction of heart disease. Furthermore, to aid in the prediction of cardiovascular diseases (SVCs), six classifiers are used: Random Forest (RF), Gradient Boosting (GB), Support Vector Classifier (SVC), Decision Tree (DT), Adaboost (AB), ensemble classifier (EC), and one type of dimensional reduction model based on principal component analysis (PCA). Grid search cross validation (GridSearchCV) is used as an automated hyper-parameter tuning algorithm. Thirty percent of the 920 patients with 14 attributes used in the Jupyter Notebook are test data. Evaluation measures that include precision, accuracy, recall, F1-score, confusion matrix, and receiver operating characteristic (ROC) curve are used to assess the performance of six classifiers.