Machine Learning for Early Detection of Type 2 Diabetes Based on Liver Enzymes and BMI

This work investigates the predictive role of liver enzymes and body mass index (BMI) in the development of type 2 diabetes. We conducted an analysis on a publicly available Chinese demographic dataset to evaluate different health parameters. The dataset was obtained from a retrospective cohort study involving 211,833 adults from 11 cities in China who were free of diabetes at baseline. The key variables examined comprised demographic information, physical measurements, blood pressure, fasting plasma glucose levels, lipid profiles, liver enzymes, lifestyle factors, and family medical history. A Logistic Regression model has been employed for early detection of type 2 using ALT, FPG, triglyceride, and cholesterol levels. The presented technique showcased its efficiency by achieving a classification accuracy of 88.9% and a recall score of 72%. The results underscore the importance of BMI as a significant risk factor for diabetes, particularly in younger age groups, and highlight the utility of predictive modeling in identifying at-risk individuals for early intervention.