Cardiovascular Disease Detection Using Machine Learning

Abstract

The detection of Cardiovascular Diseases (CVDs) prematurely is of great interest for the Healthcare Industry. According to the World Health Organization, heart diseases represent 32% of global deaths by 2019. In this work, we propose building an interpretable machine learning model to detect CVDs. For this, we use a public dataset consisting of over 320 thousand records and 279 features. We explore the performance of three well-known classifiers and we build them using hyper-parameter techniques. For interpretability, feature relevance is tested. After the experimental results, we found Random Forest to performed the best with 94% of accuracy and 81% of area under the ROC curve. We also implement an easy web application as a tool for detecting CVDs using relevant features information.