Supervised learning applied to classification and regression problems across healthcare, agriculture, finance, and business. Built and evaluated in Python using Scikit-learn and TensorFlow.
The healthcare models draw on biological domain knowledge in feature selection and result interpretation - clinical variables carry physiological interdependencies that shape which features are meaningful predictors and which are noise.
The largest project in this repository. A decision tree model trained on historical agricultural data - region, rainfall, and temperature - to forecast crop yield.
| Metric | Score |
|---|---|
| Train R² | 0.978 |
| Test R² | 0.9579 |
Minimal degradation from training to test indicates strong generalization. Full feature analysis, methodology, and evaluation outputs are documented in the project folder.
| Project | Domain | Algorithm | Goal |
|---|---|---|---|
| Crop Yield Prediction | Agriculture | Decision Tree | Forecast yield from region, rainfall, and temperature |
| Heart Disease Classification | Healthcare | Logistic Regression | Classify patient likelihood of heart disease from clinical diagnostic features |
| Diabetes Screening | Healthcare | Naive Bayes | Predict diabetes diagnosis from patient health parameters |
| Loan Repayment Prediction | Finance | Decision Tree | Predict repayment likelihood from credit history and financial behavior |
| Customer Churn Prediction | Business | Neural Network | Identify customers at risk of churn to enable proactive retention |
| Salary Prediction | Business | Linear Regression | Predict compensation from years of experience as a regression baseline |
Each model is self-contained in its own folder with a Jupyter notebook covering the full pipeline: exploratory analysis, preprocessing, train/test split, training, and evaluation with inline commentary and output visualizations.
The root directory also contains standalone NumPy, Matplotlib, and Seaborn exercises - library practice notebooks, distinct from the project work above.