Diabetes Prediction System Using Data Analysis and Machine Learning

Diabetes mellitus is one of the most prevalent chronic diseases worldwide, affecting over 537 million adults globally. Early and accurate prediction of diabetes risk can significantly reduce complications and healthcare costs. This paper presents a Diabetes Prediction System (DPS) that integrates comprehensive data analysis with a Random Forest machine learning model trained on the Pima Indians Diabetes Dataset comprising 768 patient records. Eight clinical biomarkers — plasma glucose, body mass index (BMI), blood pressure, insulin levels, skin thickness, age, number of pregnancies, and diabetes pedigree function — are used as input features. The proposed system achieves a prediction accuracy of 76–79% with a sensitivity of 81% and specificity of 74%. Comparative analysis with Logistic Regression, Naive Bayes, and Support Vector Machine demonstrates that Random Forest yields superior performance across all evaluation metrics. The system is implemented as an interactive web application enabling real-time risk assessment for healthcare practitioners and individuals. Results confirm that glucose level and BMI are the most influential predictors, contributing 24% and 18% of feature importance respectively. The findings support the utility of ensemble machine learning methods for clinical decision support in diabetes screening programs.

Keywords: Diabetes Prediction, Machine Learning, Random Forest, Data Analysis, Pima Indians Dataset, Clinical Decision Support.

M.Arthi, G.Gayathri, R.Karunya (2026). Diabetes Prediction System Using Data Analysis and Machine Learning. International Journal of Innovative Research in Multidisciplinary Education & Technology (IJIRMET), 11(3).