Amazon Product Recommendation System

This project implements multiple recommendation engine architectures to provide personalized product suggestions using the Amazon Electronics dataset. As a Data Science project, it explores the transition from simple popularity-based ranking to advanced collaborative filtering and matrix factorization techniques.

🚀 Overview

In modern e-commerce, personalized recommendations are essential for engagement. This project simulates the role of a Data Science Manager at Amazon, tasked with building a system to recommend products based on historical user rating data.

🛠️ Technical Stack

• Language: Python
• Libraries: pandas, numpy, matplotlib, seaborn
• ML Framework: scikit-surprise (for KNN and SVD models)
• Environment: Google Colab / Jupyter Notebook

📊 Models Implemented

The repository contains three distinct recommendation strategies:

1. Rank-Based Recommendation System

   • Targets the "Cold Start" problem for new users.
   • Recommends products based on highest average ratings and minimum interaction thresholds.

2. Collaborative Filtering (User-User & Item-Item)

   • User-User: Finds similar users based on their rating patterns using Cosine Similarity and KNN.
   • Item-Item: Recommends products similar to those a user has already liked.
   • Includes hyperparameter tuning via GridSearchCV to optimize $k$ neighbors and similarity metrics.

3. Model-Based Collaborative Filtering (SVD)

   • Uses Singular Value Decomposition (SVD) to identify latent features in the user-item matrix.
   • Handles sparse data more effectively than basic similarity models.

📈 Key Results

The models were evaluated using RMSE, Precision@k, Recall@k, and F1-Score:

• Baseline SVD RMSE: 0.9217
• Optimized SVD F1-Score: 0.878
• Optimized User-User F1-Score: 0.884

📂 Dataset Details

The dataset used contains over 7.8 million observations. To ensure computational efficiency and recommendation quality, the data was filtered to include:

• Users with at least 50 ratings.
• Products with at least 5 ratings.

💡 How to Use

1. Clone the repository.
2. Ensure scikit-surprise is installed: pip install surprise.
3. Open recommendation_systems_learner_notebook_full_code_-1-2.ipynb in Google Colab or Jupyter.
4. Run the cells to see exploratory data analysis and model performance comparisons.