Amazon Product Recommender System

A collaborative filtering–based recommender system built using the Amazon Electronics Reviews dataset. This project implements multiple models (User-User, Item-Item, SVD) to predict user preferences and optimize personalized product recommendations.

---

Dataset

• Source: Amazon Electronics Reviews
• Size (raw): ~7.8 million interactions
• Filtered to: Users with ≥50 ratings and items with ≥5 ratings (~65K interactions)

See data/README.md for dataset structure and download instructions.

---

Technologies Used

• Python · pandas · NumPy · matplotlib · seaborn
• Recommender Engines: surprise library — KNNBasic, SVD
• Evaluation Metrics: Precision@k, Recall@k, F1 Score, RMSE

---

Models Implemented

1. Rank-Based Recommender: Top-rated popular items
2. User-User Collaborative Filtering (KNN)
3. Item-Item Collaborative Filtering (KNN)
4. SVD Matrix Factorization

Each model is hyperparameter-tuned and evaluated on multiple metrics.

---

Evaluation

Final metrics (see results/final_metrics.txt):

• Best Recall: SVD Tuned (Recall@10 = 0.903)
• Best RMSE: SVD Tuned (0.895)
• Best Precision: User-User CF (Precision@10 = 0.856)

Visual: images/rating_plot.png shows distribution of original ratings.

---

How to Run

pip install -r requirements.txt
python src/recommender.py

---

Summary & Recommendations

See report/summary.md for business conclusions, model insights, and strategic tradeoffs.

---

Author

Eliana Gabriela Matos Polanco

• GitHub: elianagm
• Email: elianagmpolanco@gmail.com

---