Amazon Product Co-Purchasing Network Analysis

This project focuses on analyzing Amazon's co-purchasing network metadata to understand the connections between products and customer behaviors. Using graph-based techniques, we created primary networks and subgraphs to explore customer-product and product-product relationships. The project leverages publicly available datasets and tools for social network analysis to derive actionable insights.

---

Project Overview

1. Dataset

The dataset used in this project was compiled by Jure Leskovec, Lada A. Adamic, and Bernardo A. Huberman during the summer of 2006 as part of their study on viral marketing dynamics. It is available through the Stanford Network Analysis Project (SNAP). The dataset includes:

• 548,552 products, such as books, music CDs, DVDs, and VHS tapes.
• Attributes such as:
  • Title
  • Sales rank
  • Product categories
  • Similar products
  • Detailed customer reviews (time, ID, rating, votes, helpfulness).

This dataset provides the foundation for creating customer-product and product-product networks.

---

2. Key Steps

1. Primary Network Construction:

   • A primary customer-product network was constructed using the primary_network_construction.ipynb file.
   • This includes over 2.5 million nodes and 8 million edges, representing customers and products and their interactions.

2. Subgraph Creation:

   • Due to the computational complexity of analyzing the entire network, we focused on smaller subgraphs.
   • The subgraph creation process is documented in the Subgraph_Analysis.ipynb file.
   • The final subgraph contains approximately 1,000 nodes and preserves key characteristics of the primary network.

3. Reproducibility:

   • To ensure reproducibility, the generated subgraph is included in this repository (final_subgraph.pkl).

---

3. Methodology

• Customer-Product Network:

  • Nodes represent customers and products.
  • Edges connect customers to products they reviewed or purchased.
  • Edge attributes include review weight, rating, and timestamps.

• Product-Product Network:

  • Nodes represent products.
  • Edges connect products identified as "similar" in the dataset.
  • Attributes include average product ratings.

• Subgraph Construction:

  • A representative subgraph was extracted using breadth-first search (BFS) from seed nodes selected based on degree centrality and product categories.
  • The subgraph preserves key properties such as community structures and local dynamics.

---

4. Tools and Libraries

The project was developed using Python and the following libraries:

• NetworkX: For graph creation and analysis.
• Pandas: For data preprocessing and tabular analysis.
• Scikit-learn: For clustering algorithms.
• Matplotlib/Seaborn: For data visualization.
• Community Detection Package: For community-based analyses.

---

Citation

If you use this dataset or methodology, please cite:

• J. Leskovec, L. Adamic, and B. Huberman: The Dynamics of Viral Marketing. ACM Transactions on the Web (TWEB), 1(1), 2007.
• Dataset: Amazon Meta-Data

---