Personalized lifestyle recommendations for improved diabetes management leveraging machine learning

Abstract

Diabetes is an extremely dangerous condition that is rapidly expanding, and its early diagnosis and effective management are crucial. Healthcare professionals must prioritize rapid diagnosis and personalized treatment strategies; however, with increasing patient numbers, existing healthcare facilities may be unable to meet this increased demand. As a result, it is critical that patients adopt self-management practices that are easy to comprehend using machine learning techniques. In this study, real-time blood glucose levels, blood pressure, and other lifestyle factors such as diet, exercise, stress, and sleep were collected, and Exploratory Data Analysis was conducted to compare the findings with those of previous research. Caloric estimation was performed using an extensive formulation for each participant. The parameters were divided into nine distinct cluster models using the K means clustering technique, and the resulting clusters were examined for features that yielded significant observations. The classification algorithms were used to segregate the new data into their appropriate clusters, and the obtained data were subjected to 5-fold cross validation to avoid overfitting and were classified into clusters using different classification algorithms such as Random Forest, Decision Tree, Naive Bayes, Logistic Regression, Support Vector Machine, and K-Nearest Neighbor. Pertinent recommendations were provided to each cluster’s members based on the literature. An interactive web application was created by integrating machine learning models to improve user experience, which could be integrated into wearables in the future to revolutionize healthcare.