MaxPWSCan

Implementation of “Graph-based Substructure Pattern Mining with Edge-Weight”

Reference Paper:
Md Ashraful Islam, Chowdhury Farhan Ahmed, Md Tanvir Alam & Carson Kai-Sang Leung.
Graph-based substructure pattern mining with edge-weight. Applied Intelligence, Vol. 54 (5), pp. 3756-3785 (2024).
📄 Read the Paper on Springer :contentReference[oaicite:2]{index=2}

---

🎯 Overview

This repository contains the implementation of the algorithm and experimental framework described in the above paper.
The method mines frequent substructures from a database of edge-weighted graphs — capturing not just structural frequency but also the importance via edge weights.
A key contribution is the MaxPWS pruning technique which supports mining weighted frequent subgraphs while maintaining completeness despite the lack of the downward-closure property in the weighted setting. :contentReference[oaicite:3]{index=3}

---

🗂️ Repository Structure

├── src/                   → core algorithm implementation  
├── data/                  → input graph database / example graph files  
├── results/               → output files, pattern lists, evaluation metrics  
├── scripts/               → helper scripts for preprocessing or launching experiments  
├── README.md              → this file  
└── LICENSE                → (if applicable) license file  

---

⚙️ Requirements

• Java / Scala / Python (depending on your implementation)
• Graph library dependencies (list here: e.g., networkx, igraph, etc.)
• Java 8+ or equivalent if you use a Java-based tool

---

🚀 Usage

1️⃣ Prepare your graph dataset

Place your collection of edge-weighted graphs into data/.

2️⃣ Launch the mining algorithm

# Example: run the main mining script
java -jar MaxPWSCan.jar --input data/graphdb/ --min_sup 0.05 --min_weight 10

(Adjust the command to match your actual script/implementation)

3️⃣ Review results

The discovered subgraph patterns and evaluation metrics will be saved under results/. Inspect runtime logs, pattern counts, and graphs.

---

📊 Experimental Findings

According to the paper, the proposed framework:

• Significantly reduces search space via MaxPWS pruning. :contentReference[oaicite:4]{index=4}
• Maintains completeness of mining while handling weighted support calculation. :contentReference[oaicite:5]{index=5}
• Achieves faster runtime compared to prior methods for static and dynamic edge-weight settings. :contentReference[oaicite:6]{index=6}

---

📘 Citation

If you use this code in your research or projects, please cite the original publication:

Islam, M.A., Ahmed, C.F., Alam, M.T., & Leung, C.K. (2024). Graph-based substructure pattern mining with edge-weight. Applied Intelligence, 54(5), 3756-3785. https://doi.org/10.1007/s10489-024-05356-7

---

🤝 Contributing

Contributions, bug reports, and enhancements are welcome! Please submit pull requests and describe your changes clearly.

---

📄 License

(Insert your license here – e.g., MIT, Apache 2.0 or other.)

---

📧 Contact

Md Ashraful Islam — [email address]
GitHub: cseduashraful