A collection of my machine learning projects, assignments, and experiments built throughout my ML learning journey. This portfolio is meant to showcase my growing skills and serve as a reference for anyone interested in exploring or reusing the work.
I'm actively learning machine learning and documenting that journey here - from foundational concepts to hands-on projects. Each project includes a data exploration or data exploration with model building, and accompanying key takeaways.
| Project | Description | Topics | My Learning |
|---|---|---|---|
| CS6140 - Data Exploration (Movielens) | Exploration and analysis of the MovieLens dataset | EDA, Collaborative Filtering | — |
| CS6140 - Data Pipeline | End-to-end data preprocessing pipeline on a customer dataset | Missing Value Imputation, Encoding, Scaling, sklearn Pipeline | Understood data leakage, why scalers must fit on train only, when to use OneHot vs Label encoding, and how sklearn Pipeline prevents manual errors |
| CS6140 - Principal Component Analysis (Wine) | PCA on Wine dataset — dimensionality reduction, variance analysis, and classifier comparison | PCA, Logistic Regression, SVC, KNN, Decision Boundaries, Loadings Heatmap | Understood that variance explained ≠ classification accuracy — only 3 PCs needed for 100% accuracy vs 10 PCs for 95% variance. SVC with RBF kernel was most consistent classifier. |
| CS6140 - Association Rule Mining (Market Basket) | Apriori-based market basket analysis — from-scratch preprocessing, parameter sensitivity experiments, and a rule visualization dashboard | Apriori, Support/Confidence/Lift, Parameter Tuning, Custom Rule Scoring, Data Visualization | Learned that lift (not frequency) reveals meaningful associations — common items dilute their own lift by co-occurring with everything, while niche item pairs produce the strongest, most surprising rules. |
- Python
- Jupyter Notebooks
- Pandas, NumPy
- Scikit-learn
- Matplotlib / Seaborn
This portfolio is a work in progress - new projects added regularly as I learn.