Machine Learning-Based Approach for Diabetes Detection and Prediction

Diabetes, is a group of metabolic disorders in which there are high blood sugar levels over a prolonged period. If it was possible to predict the chance of the development of the disease then more timely intervention might help reduce the extent to which the disease effects its victims and health care services. This study focuses on the use of Machine Learning (ML) algorithms in analyzing patient data for the purpose of predicting diabetes risk. The Logistic Regression, Decision Tree, K-Nearest neighbors, Random Forest, Gradient Boost, Naive Bayes and XGBoost algorithms were employed to classify diabetes status, leveraging demographic and clinical features from a dataset of patient records. Each model’s effectiveness was assessed through preprocessing steps such as encoding categorical variables (e.g., gender and smoking history) and scaling numerical attributes (e.g., age, BMI, blood glucose levels). These classifiers have been evaluated using metrics which included accuracy, precision, recall and F1-score to gain the predictive performance and generalizability. Among all the models, Gradient Boost and XGBoost classifiers demonstrated the highest accuracy having 86% and 0.85 and robustness which makes them highly suitable for clinical decision-making applications in diabetes diagnosis and management.