Machine Learning Model in Diagnosing Diabetes Mellitus among Women of Reproductive Age Using Kaggle Datasets

Abstract

The diagnosis of diabetes mellitus using machine learning (ML) offers a modern, efficient, and non-invasive approach to identifying the disease. ML algorithms analyze complex patterns in patient data, such as blood glucose levels, demographic factors, lifestyle habits, and genetic information, to make accurate predictions. This study aimed to develop and implement a machine-learning model for predicting diabetes based on clinical and demographic data such as glucose levels, blood pressure, BMI, and Diabetic predegree Function. The data consists of 768 females at least 21 years old. The datasets consist of several medical predictor variables and one target variable, Outcome. Predictor variables include the number of pregnancies, BMI, blood glucose, skin thickness, Diabetes Predegree Function, insulin level and age. The study adopted the support vector machines (SVM) for classification. The data was divided into training sets (70%) and testing (30%), both of which were trained. The model was evaluated for performance using precision and accuracy scores. Finding reveal that there are 500 (65%) diabetics and 268 (35%) non-diabetics. The mean values are Pregnancies: 3.85 ± 3.37. Glucose: 69.11 ± 19.36 mg/dL. Blood Pressure: 20.54 ± 15.95 mmHg. Skin Thickness: 20.54 ± 15.95 mm. Insulin: 79.80 ± 115.24 μU/mL. BMI: 31.99 ± 7.88. Diabetes Pedigree Function: 0.47 ± 0.33. Age: 33.24 ± 11.76 years. The accuracy of this model in predicting diabetes for the training data was 88.66 % and 87.27% for the test data set. Overall, the performance of the machine learning model presented in this study indicates that this approach can be effective for diabetes prediction in clinical practices.