This is for KDD'24-submission:
Dual-Channel Multiplex Graph Neural Networks for Recommendation
Recent versions of the following packages for Python 3 are required:
- numpy==1.21.2
- torch==1.9.1 + cuda11.1
- scipy==1.7.1
- scikit-learn==0.24.2
- pandas==0.25.0
- tqdm==4.61.2
All of the datasets we use are publicly available datasets.
The used datasets are available at:
- Retail_Rocket https://tianchi.aliyun.com/competition/entrance/231719/information/
- Tmall https://tianchi.aliyun.com/dataset/dataDetail?dataId=649
- Yelp https://www.yelp.com/dataset/download
We compress the data set into a mat format file, which includes the following contents.
- edges: array of subnetworks after coupling, each element in the array is a subnetwork.
- features: attributes of each node (user\item) in the network.
- labels: label of labeled points.
- train: index of training set points.
- valid: index of validation set points.
- test: index of test set points.
In addition, we also sample the positive and negative edges in the network, and divide them into training, validation and test fields for recommendations.
First, you need to choose the specific dataset for recommendation tasks in Recommendation.py. Second, you need to modify the number of weights in Model.py. The number of weights should be the number of sub-networks after decoupling. Finally, you need to determine the sub-network and the number of sub-networks in Decoupling_matrix_aggregation.py. Commonly hyperparameters can be reset in args.py. relation_chain.py and rc_aware_encoder.py construct the relation chains and the relation chain-aware encoder in DCMGNN that captures the relation sequence, and the correlations and dependencies between different auxiliary relations and the target relations, respectively.
Execute the following command to run the recommendation task:
python Recommendation.py