LLM-Enhanced Heterogeneous Graph Neural Networks for Research Project Conflict Detection

Environment Setup

Requirements

• Python 3.8+
• PyTorch 1.12+
• DGL 1.0+
• scikit-learn
• numpy
• pandas

You can install the required packages using:

pip install torch dgl scikit-learn numpy pandas

Running the Code

Basic Usage

To train the model with default parameters:

python main.py

Main Parameters

• --seed: Random seed (default: 12)
• --n-hidden: Number of hidden units (default: 16)
• --n-layers: Number of GNN layers (default: 2)
• --dropout: Dropout rate (default: 0.1)
• --lr: Learning rate (default: 1e-3)
• --n-epochs: Number of training epochs (default: 1000)
• --trials: Number of trials (default: 10)
• --use_attention: Enable attention mechanism (0/1)
• --use_contrast: Enable contrastive learning (0/1)
• --use_llm: LLM enhancement type ('none'/'type0'/'type1')
• --lambda_contrast: Weight for contrastive loss (default: 0.05)
• --temperature: Temperature for contrastive loss (default: 0.5)

Output

The model will output:

• Training process metrics
• Final test results including AUC-ROC, AUC-PR, and Recall@K
• Results will be saved to CSV file (default: 'results/ours_results_tmp.csv')

Data

Place your data files in the processed_data directory:

• hetero_graph_woLLM_split.pth: Graph without LLM enhancement
• hetero_graph_wLLM_type0_split.pth: Graph with Type-0 LLM enhancement
• hetero_graph_wLLM_type1_split.pth: Graph with Type-1 LLM enhancement