GlucoSense

The problem

Diabetes is often detected late, when intervention is harder and costlier.

What I built

A scikit-learn pipeline with careful feature selection and median imputation, comparing multiple classifiers on precision, recall, F1 and AUC.

Highlights

• Extra Trees, XGBoost and SVM compared on an imbalanced dataset
• Correlation-based feature pruning to find the signals that matter
• Completed as an AI/ML intern at Infosys Springboard

How it works

1. 01Health & lifestyle data
2. 02Median imputation
3. 03Feature pruningCorrelation-based
4. 04Model comparison
5. 05EvaluationPrecision · recall · F1 · AUC