Design and Implementation of an Artificial Intelligence-based Mobile Application for Diabetes Risk Prediction

Abstract

The rising global prevalence of diabetes, coupled with its severe complications and the burden it places on healthcare systems, has created an urgent need for innovative solutions that can provide early risk detection and preventive intervention. Traditional diagnostic methods are often reactive, relying on clinical tests conducted after symptoms appear, which limits their effectiveness in minimizing disease progression. This paper presents an artificial Intelligence-based mobile application for health risk prediction, specifically focusing on diabetes mellitus. The mobile application is implemented to enable individuals and healthcare providers to access accurate and proactive health risk assessment. The mobile application incorporates relevant individual health information and lifestyle data, such as medical records, demographic information, and behavioral factors, into the predictive model, and evaluates the system using a probability risk score. The system integrates a module that enables users to register and authenticate their details, input their personal and behavioral information required to develop the artificial intelligence model, predict diabetes risk score, and recommend appropriate lifestyle changes to minimize risk factors. The system adopted Object-Oriented Software Development Methodology and data modelling approach encompassing data preprocessing, feature selection, and learning analytics, and a mobile application was developed in Python and Next.js. The Behavioral Risk Factor Surveillance System (BRFSS) dataset was used for model training using the Light Gradient Boosting model (Light GBM. Results showed that Light GBM achieved an optimal performance, producing a precise risk score. The findings confirmed that AI-based systems can effectively support preventive medical intervention, offering a scalable and practical tool for modern healthcare.