REDICTING EARLY DIABETES MELLITUS WITH MACHINE LEARNING TECHNIQUES AT ADARE GENERAL HOSPITAL,

Abstract

Diabetes mellitus is a chronic metabolic disorder that affects a large proportion of the global population. About 422 million people worldwide have diabetes, the majority living in low-and middle-income countries, and 1.5 million deaths are directly attributed to diabetes each year. Both the number of cases and the prevalence of diabetes have been steadily increasing over the past few decades according to World Health Organization (WHO) Global Report on Diabetes 2016. According to the IDF Diabetes Atlas 10th Edition, an estimated 24 million adults aged 20-79 years were living with diabetes in the IDF Africa Region in 2021, representing a regional prevalence of 4.5%. 54% of people living with diabetes in the region are undiagnosed, the highest proportion of all IDF Regions. It is also predicted that the total number of people with diabetes will increase by 129% to 55 million by 2045. As per the data provide by IDF, Ethiopia has total adult population of 57,503,700, out of this, prevalence of diabetes in adults is 3.3%, while total cases of diabetes in adults in Ethiopia is 1,920,000[7]. Early detection of diabetes is critical for successful management and prevention of long-term complications. Machine learning (ML) has emerged as a promising approach for the early prediction of diabetes. In this research, we developed an ML- based diabetes mellitus prediction model using a dataset of patient clinical data from Adare General Hospital, including age, gender, body mass index (BMI), Glucose, blood pressure, and various blood test results. In this research, we used the popular ML algorithms, including K-Nearest Neighbor (KNN), Decision Trees (DT), Support Vector Machine (SVM), Artificial Neural Network (ANN), Logistic Regression, Naive Bayes and Random Forest classifier for early diabetes mellitus prediction. Our results showed that the random forest algorithm outperformed other methods, achieving an accuracy of 97.4% in predicting diabetes. Additionally, our research identified blood glucose, age, BMI levels, blood pressure, history of hypertension, polyphagia and polydipsia as the most significant predictors of diabetes. Our findings suggest that ML-based models can be a useful tool in the early detection of diabetes and have the potential to improve patient outcomes by enabling timely interventions and prevention strategies