Problem. Predict diabetes onset from clinical features to support early intervention.

Data. ./diabetes_clean.csv (columns include: pregnancies, glucose, diastolic, triceps, insulin, bmi, dpf, age, diabetes).

Approach.

• Baseline Logistic Regression; ROC curve + confusion matrix.
• Compared against KNN (logistic performed better across metrics).
• Hyperparameter tuning with GridSearchCV and RandomizedSearchCV.
• Evaluated with train/test split; tracked accuracy, precision, recall, F1, ROC-AUC.

Results.

• Logistic > KNN on all reported metrics.
• ROC-AUC ≈ 0.801; accuracy ≈ 0.68; balanced performance across classes.

What I Learned.

• Interpreting coefficients and thresholds via ROC.
• Why CV-based tuning improves generalization.
• How metric choice (F1 vs AUC) shifts model selection.

# clone if standalone
git clone https://github.com/Joe-Naz01/fine_tuning_supervised.git
cd fine_tuning_supervised

python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt
jupyter notebook