Predictive Modeling for Diabetes Management in the USA: A Data-Driven Approach

Abstract

Diabetes, especially Type 2 diabetes, has emerged as one of the major chronic conditions in the United States, affecting millions and with significant risks to public health. Coupled with this rise in prevalence is the dramatic rise in healthcare costs associated with the disease. The prime objective of this research project was to establish how predictive modeling can be used to enhance the management and prevention of diabetes in the United States. This study focused on the deployment of predictive modeling methods to support diabetes management in the United States, with an emphasis on data-driven decision-making in clinical settings and public health policy. The dataset for this research project was retrieved from accredited and credible dataset sources. The Diabetes prediction dataset included medical and demographic data of the patients along with their respective diabetic status. The provided data included age, gender, body mass index, hypertension, heart disease, smoking history, HbA1c level, and blood glucose level. In this work, the models used were Logistic Regression, Random Forest, and Support Vector Classifiers. Random Forest outperformed other models in all metrics with the highest accuracy, precision, recall, and F1-score scores. SVM had a slightly lower performance than Random Forest but still outperformed Logistic Regression in all metrics. Overall, the Random Forest was the most effective model on this particular dataset, followed by SVM and Logistic Regression. Predictive modeling can bring potential transformation to diabetes management and prevention, furnishing health professionals with actionable insights to enable improved patient outcomes in the USA. Integration of predictive models into clinical workflows may further simplify diabetes care. For instance, predictive algorithms can be integrated into EHR systems to flag patients for closer monitoring or further testing.