Abstract

Diabetes is a major global health concern that can lead to severe complications if not detected early. Type 2 diabetes often grows progressively, with many individuals unaware of their condition until significant health damage occurs. Early detection is therefore essential to enable timely medical intervention and lifestyle changes to improve treatment outcomes and reduce long-term risks. In this study, significant health factors, including insulin, blood pressure, age, glucose levels, BMI, and family history, are analysed to predict diabetes in early stages by applying machine learning (ML) models. Numerous techniques, including Logistic Regression (LR), K-Nearest Neighbors (KNN), Decision Trees (DT), Support Vector Machines (SVM), Random Forest (RF), and Extreme Gradient Boosting (XG Boost), have been employed for estimation. A hybrid model combining Principal Component Analysis, Random Forest, and XG Boost was developed to enhance prediction accuracy. Principal Component Analysis selects the features, Random Forest ensures model stability and reduces overfitting, while XG Boost improves performance by minimising prediction errors. The proposed hybrid model has achieved a higher accuracy of 99.25%, surpassing the other models included in our study. The model also proved strong recall and precision, confirming its effectiveness in categorising diabetic and non-diabetic cases. These results confirm that the proposed model can serve as an effective and reliable tool for early detection of diabetes and help healthcare professionals.

Authors

B. Mouli Chandra¹, K. Sudharshan², C.H. Silpa³, Jeyaprabhavathi Perumal⁴, D.V. Ashok⁵, Kommu Kishore Babu⁶, M. Mohana Prasad⁷
QIS College of Engineering and Technology, India^1,2,3,5,7, Arden University, United Kingdom⁴, Vignan’s Foundation for Science, Technology and Research, India⁶

Keywords

Type 2 Diabetes, Early Prediction, Machine Learning, Random Forest, XG Boost, Hybrid Model