Graph-Based Recommendation System for E-commerce using FAISS

Business Context

With the vast number of products available on eCommerce platforms, personalizing content for users has become crucial to enhance engagement and customer satisfaction. Recommendation systems play a vital role in achieving this personalization. While traditional recommendation systems like Content-Based Filtering and Collaborative Filtering are widely used, graph-based approaches have gained significant traction in recent years due to their ability to model complex relationships between users and items effectively.

This project focuses on building a graph-based recommendation system by generating graph embeddings and utilizing similarity search with FAISS (Facebook AI Similarity Search) to deliver personalized product recommendations for an eCommerce platform.

---

Aim

To build a graph-based recommender system that recommends the best products to users on eCommerce platforms based on their purchase and search history.

---

Data Overview

The dataset contains user information with nine attributes collected from an eCommerce website. The attributes include:

• User ID
• Product ID
• Purchase history
• Search queries
• Product categories
• Time of interaction
• Purchase frequency

Note: Embedding data is not pushed to the repository due to upload size limitations.

---

Tech Stack

• Programming Language: Python
• Libraries and Frameworks:
  • pandas and numpy for data manipulation and analysis.
  • pecanpy and gensim for graph embeddings.
  • umap for dimensionality reduction.
  • faiss for similarity search and recommendations.
  • plotly for interactive visualizations.
  • DuckDB for database management.
• File Management: Parquet format for efficient storage and processing.
• Database Management: SQL querying for data manipulation.

---

Approach

1. Problem Understanding

• Define the scope of the recommendation system.
• Analyze the data to identify user-item interaction patterns.

2. Graph Construction

• Use interaction data to construct user-product graphs.
• Save the graphs in .edg format for compatibility with embedding tools.

3. Embedding Generation

• Generate random walks on the graph using techniques like DeepWalk and Node2Vec.
• Train embedding models using Word2Vec and Node2Vec algorithms.
• Save the embeddings in Parquet format for further use.

4. Dimensionality Reduction and Visualization

• Use UMAP (Uniform Manifold Approximation and Projection) for reducing embedding dimensions.
• Visualize user-product clusters to identify patterns and similarities.

5. Recommendation System Development

• Use FAISS (Facebook AI Similarity Search) to build a scalable recommendation engine.
• Perform similarity searches on the embedding vectors to recommend products.

6. Result Analysis

• Evaluate the quality of recommendations.
• Visualize clusters and analyze user preferences.

---

Project Structure

.
├── data/                                  # Contains data files.
│   ├── ConstructedGraph/                  # Generated graphs from user-item interactions.
│   ├── Edg_Graphs_DataFile/               # Graphs saved in .edg format.
│   ├── Embedding_Data/                    # Generated graph embeddings.
├── documentation/                         # Learning resources and reference materials.
├── models/                                # Saved models for embeddings and recommendations.
├── notebooks/                             # Jupyter notebooks for experimentation.
│   ├── Data_Exploration_and_Analysis.ipynb
│   ├── Graph_Construct.ipynb
│   ├── Deepwalk_and_Node2Vec_Training.ipynb
│   ├── Result_Analysis.ipynb
│   ├── Embedding_Vector_Search_with_FAISS.ipynb
├── scripts/                               # Python scripts for modular tasks.
├── requirements.txt                       # List of dependencies with versions.
└── README.md                              # Project documentation.

---

Getting Started

1. Clone the Repository

git clone <repository_url>
cd <repository_folder>

2. Install Dependencies

Install the required libraries using the following command:

pip install -r requirements.txt

3. Run the Project

• Train embedding models and generate recommendations:

python scripts/train_embeddings.py

• Perform similarity search using FAISS:

python scripts/faiss_search.py

4. Explore Results

• Visualize user-item clusters and analyze results using the Jupyter notebooks in the notebooks folder.
• Check the Embedding_Data folder for saved embeddings and recommendations.

---

Results

• Personalized Recommendations:
  • Generated product recommendations based on user purchase and search history.
• Graph Insights:
  • Visualized clusters to understand user-product relationships.
• Efficient Similarity Search:
  • FAISS enabled fast and scalable recommendations.

---

Contributing

Contributions are welcome! To contribute:

1. Fork the repository.
2. Create a feature branch:

git checkout -b feature-name

3. Commit your changes:

git commit -m "Add feature"

4. Push your branch:

git push origin feature-name

5. Open a pull request.

---

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

Contact

For any questions or suggestions, please reach out to:

• Name: Abhinav Navneet
• Email: [email protected]
• GitHub: AjNavneet

---

Acknowledgments

Special thanks to:

• FAISS for enabling fast and efficient similarity searches.
• pecanpy and gensim for graph embeddings.
• umap for dimensionality reduction.
• DuckDB for its fast and efficient database capabilities.

---