A Machine Learning-Based Approach for Cardiovascular Diseases Prediction

Abstract

In recent decades, the development of technology and the increase of living standards have affected people's attention in healthcare. The healthcare market has expanded along with great attention from researchers, and the study of disease forecasting has become an inevitable process in future healthcare development. Cardiovascular disease is a category for the type of disease that causes a negative impact on the heart and blood vessels. As one of the most common and lethal potential diseases, predicting cardiovascular diseases helped in high-risk patients' decision-making process, which identifies the potential threats in the early stages. With the help of tremendous medical data and increased computational capabilities, machine learning has proved one of the most efficient prediction methods in cardiovascular disease forecasting. However, most of the proposed research either focuses on single algorithms with different parameter settings or large-scale selected algorithms with single criteria for modeling. This experiment aims to study the performance of small-scale selected algorithms with multiple criteria used in the modeling process. Specifically, this study used historical healthcare data with fourteen attributes selected. The experiment results show the Random Forest built by Classification and Regression Trees (CART) has dominant performance among all selected algorithms, which can help construct an alert system for cardiovascular disease prediction.