MACHINE LEARNING ALGORITHMS FOR EARLY DETECTION OF DIABETES USING MEDICAL RECORDS: A CASE STUDY OF UNIVERSITY COLLEGE HOSPITAL, IBADAN

ABSTRACT Diabetes mellitus is a major public health concern in Nigeria, where late diagnosis, inadequate diagnostic infrastructure, and high screening costs continue to hinder effective disease management. Early detection remains crucial for reducing complications and improving patient outcomes. To address this challenge, this study developed and evaluated machine learning algorithms capable of predicting diabetes using anonymized clinical records obtained from the University College Hospital (UCH), Ibadan. The study adopted the Design Science Research Methodology (DSRM) to guide the design, development, and validation of the predictive system. Three supervised machine learning models—Logistic Regression, Random Forest, and Support Vector Machine—were trained using key clinical features such as glucose level, age, BMI, blood pressure, gender, family history, and HbA1c. The dataset underwent rigorous preprocessing, including missing-value imputation, normalization, outlier treatment, feature engineering, and class balancing using the Synthetic Minority Oversampling Technique (SMOTE). These processes enhanced data quality and ensured robust model training. Model performance was evaluated using accuracy, precision, recall, F1-score, and ROC-AUC metrics. Among the implemented models, the Random Forest classifier demonstrated superior predictive performance, showing strong discriminative ability and effective generalization to unseen data. A functional prototype decision-support system was developed using the best-performing model to demonstrate real-world clinical applicability and support healthcare professionals in early diabetes screening. Overall, the study demonstrates that machine learning provides a reliable, scalable, and cost-effective approach to strengthening early diabetes diagnosis in Nigeria. The findings highlight the transformative potential of artificial intelligence to improve data-driven healthcare delivery, especially in environments with limited diagnostic capacity. Keywords: Diabetes Mellitus; Machine Learning; Random Forest; Logistic Regression; Support Vector Machine; Early Detection; Clinical Decision Support System; Medical Records; Predictive Analytics; UCH Ibadan.