Smart Diabetes: An Intelligent Classifier for Continuous Health Monitoring

Abstract

Diabetes, a chronic and widespread condition, needs precise and effective management strategies to mitigate its impact on patients' lives. This paper explores diabetes care with a focus on classification, such as whether the patient is diabetic or not. This is done by applying machine learning models to enhance management using Kaggle's “Diabetes Prediction Dataset”. The paper begins with precise data preprocessing, setting the stage for evaluating various models. The Random Forest model emerges as the frontrunner, demonstrating superior accuracy with metrics such as accuracy of 89.160%, precision of 87.620%, recall of 88.304 %, F1 Score of 87.961%, and AUC ROC Score of 96.820%. This analysis not only advances academic insights into machine learning's role in healthcare but also has tangible implications for real-world diabetes management, paving the way for ongoing health monitoring and marking a significant stride towards more effective diabetes care.