Open-GNN-Mini-Competition-

GNN-based Fake News Detection Challenge

Welcome to the GNN-based Fake News Detection Challenge! This competition focuses on detecting fake news propagation on Twitter using Graph Neural Networks (GNNs). The task is based on the User Preference-aware Fake News Detection (UPFD) framework. Live Leaderboard

Participants are asked to improve the baseline GNN model by including user profile features in addition to existing text embeddings.

---

Repository Structure

Open-GNN-Mini-Competition/
├── data/
│   ├── gossipcop
│   ├── metadata
│   ├── 
│   ├── 
│   ├── 
│   └── 
├── submissions/
│   └── sample_submission.csv
├── models/
│   ├── saved_model.model
│   └── 
├── scoring_script.py
├── dataloader.py
├── evaluate.py
├── leaderboard.html
├── leaderboard.json
├── model.py
├── test.py
├── train.py
├── requirements.txt
├── README.md
└── LICENSE

---

🗂 Dataset

We use the GossipCop and Politifact datasets, which contain Twitter news propagation graphs. Each graph represents a news article as the root node, and the users who retweeted the news as child nodes. You can download the data from this link : https://drive.google.com/drive/folders/1OslTX91kLEYIi2WBnwuFtXsVz5SS_XeR

Node Features

• Text embeddings of the news (root node) and historical tweets of users
  • Pretrained spaCy word2vec (300-dim) or BERT embeddings (768-dim)
• Task extension: include user profile features (10-dim):
  • Account age
  • Verified status
  • Number of followers/friends
  • Number of tweets
  • Geolocation enabled
  • Description length
  • etc.

Graph Labels

• 0: Real news
• 1: Fake news

---

📝 Problem Statement

Task: Classify each news propagation graph as real or fake.

• Baseline: Uses only text embeddings of the news and historical tweets of users.
• Challenge: Improve the baseline by including user profile features in the final node embeddings.

---

⚡ Baseline Model

The baseline GNN is implemented in model.py. It supports:

• Graph Convolutional Network (GCN)
• Graph Attention Network (GAT)
• GraphSAGE

Features used in baseline:

• Text embeddings of news and historical user tweets
  Features to add for challenge:
• User profile features (10-dimensional)

---

🚀 Getting Started

1. Clone the repo

git clone https://github.com/TugaAhmed/Open-GNN-Mini-Competition-.git
cd gnn-challenge

2. Install dependencies

pip install -r requirements.txt

3. Download dataset from the link above and place the unzpied data inside data folder

4. After creating your model, run test.py that will generate sumbission.csv

Submission Workflow

• Fork the repo and add your submission CSV under submissions/
• Create a pull request
• GitHub Actions will automatically run scoring_script.py to evaluate your submission and update the leaderboard