Predict2Protect: Machine Learning Web Application in Early Detection of Heart Disease.

Abstract

Across the world, there are few universal scenarios, but the pain of losing a loved one to heart disease is an exception and a reality shared by millions every year. Heart disease is the greatest killer in society today, and one prevalent root of this issue is untimely diagnosis, often caused by unsustainable costs and lack of accessible healthcare for underserved populations. Recognizing these disparities, the goal of this project was to create an easily available application and interface for all that accurately indicates one's risk of heart disease. To address this, a machine learning model, Predict2Protect, was built in Python. An open-source dataset compiled of 1025 patients of diverse backgrounds was scaled, adjusted to include inquiries answerable by patients, and split into 75% for training, 15% for validation, and 25% for testing. Four models were tested with the hypothesis that if the RandomForestClassifier was used, it would have the highest validity. This was not supported, as the DecisionTree model had a 100% accuracy for training data and 95% for test data. Through the application software Streamlit, this program was processed into a web application that is now found in browser extensions. The application reports the risk of one having heart disease with a 95% accuracy and describes the risk percentage of developing heart disease within the next year. With a simple interface and high accuracy, Predict2Protect aims to provide a view into one's health with the goals of accessible heart disease prediction and early treatment for patients around the world.