Performance Evaluation of Machine Learning Models for Cardiovascular Disease Prediction

Cardiovascular disease (CVD) is a deadly health issue that requires urgent attention due to the increasing global population. Using technological solutions, Machine Learning (ML) could be used to tackle health challenges in medical institutions. However, CVD diagnosis often involves multiple diagnostic procedures, leading to high medical errors. Using high-frequency and gamma rays in diagnostic devices exposes patients to high risks of other diseases. The acquisition of these devices is challenging in Low and Middle-Income Countries (LMICs), where patients are less privileged, and families lose their lives due to the inability to afford them. Healthcare institutions are utilizing the adoption of Artificial Intelligence (AI) to provide innovative solutions to these issues. In this research, the dataset was obtained from Kaggle. The Cross Industry Standard Process for Data Mining (CRISP-DM) framework was used. We selected five phases excluding the Deployment phase from the framework. Then, experiments conducted starting with exploratory data analysis, data cleansing, data visualization, transforming the dataset then splitting the CVD dataset, allocating 80% of the dataset to training and the remaining 20% as a testing dataset. We applied the different ML algorithms, where we achieved the best accuracy from Random Forest, Extra Tree Classifier, and Decision Tree all with 98.18% on the training dataset, for further experimental selection we tested the dataset where we obtained the following percentages: Random Forest 79.22%, Extra Tree Classifier 78.68% and Decision Tree 73.48%. Finally, the experiment featured the Random Forest algorithm as the best classifier compared to the rest as mentioned above, as it is more robust with the ability to handle non-linear and complex relationships making it more effective for CVD models.