Empowering Diabetes Patients by Providing Machine Learning-Driven Predictions and Personalized Visualization Results

Abstract

Although millions of patients have diabetes, it is often challenging to interpret symptoms that historically lead to the condition. To solve this disparity, we created an end-to-end platform that uses a Random Forest model that predicts early-stage diabetes with 95.6% accuracy, then visualizes patient data for those with similar symptoms. After users enter their data for the five most strongly-correlated diabetes symptoms, the model predicts whether the user has diabetes. As a result, this project transforms how patients communicate about their own data, thereby serving as a mechanism to start important conversations with their doctors or others around the world.