Diabetes Prevalence Prediction Among Academic Staff of Tertiary Institutions in South-Western Nigeria using Machine Learning Techniques

Diabetes mellitus (DM) is a chronic health condition characterized by inadequate insulin production or ineffective utilization, which leads to elevated blood sugar levels. In Nigeria, the prevalence of DM has seen a sharp rise, particularly among individuals aged 20-79, with significant increases projected over the coming decades. Among academic staff in Southwestern Nigeria, the high demands of their professional duties have negatively impacted their health, leading to increased susceptibility to diabetes. This study seeks to address the limitations of existing diabetes prediction models, which primarily rely on secondary datasets, by utilizing primary data collected from academic staff in Southwestern Nigeria. A comprehensive diabetes prediction model is formulated using machine learning and ensemble methods such as K-Nearest Neighbors (KNN), Random Forest, and Logistic Regression. By employing feature selection techniques and model validation methods, the study offers novel insights into diabetes risk factors among academic staff. The results demonstrate that ensemble models, particularly Voting and AdaBoost, consistently outperformed individual machine learning algorithms, showcasing their potential for accurate diabetes prediction. This study provides a tailored and context-specific approach to diabetes prediction, with implications for public health interventions targeting tertiary institutions.