Diabetes Risk Prediction Based on Clinical and Lifestyle Data

With the continuous increase in global diabetes incidence, the use of integrated health data for early detection of risk and timely intervention has attracted growing attention in public health. Based on the National Health and Nutrition Examination Survey (NHANES), this study combines demographic characteristics, clinical test results, laboratory indicators, and lifestyle patterns to construct a diabetes risk prediction model using a Multi-Layer Perceptron (MLP). Key physiological markers— including Glucose, HbA1c, BMI, and Insulin—were analysed together with lifestyle-related variables such as dietary intake, physical activity, and alcohol use. The developed model attains an accuracy of about 85% and an AUC of 0.89 on the test dataset. Findings suggest that biomarkers associated with blood glucose regulation play a central role in predicting diabetes, while healthier diet structure and regular exercise contribute to reducing individual risk. Moreover, Principal Component Analysis (PCA) uncovers complementary associations between clinical measures and lifestyle features, improving the interpretability of the model. Overall, the study demonstrates an effective data-driven strategy for early diabetes risk assessment and personalised health management, while providing insights into the application of artificial intelligence techniques in public health research.