h2seqrec

Paper

Hyperbolic Hypergraphs for Sequential Recommendation
Data and HGCN code

Environment

Recommendation Environment

python==3.6.12
pytorch==1.9.0
numpy==1.17.5
scikit-learn==0.24.2
scipy==1.5.4 

Hyperbolic Hypergraph Neural Network Environment

pytorch==1.0.0
python==3.6.13

Run steps

1. Pre-training phase (three contrastive pre-training tasks)
   a. Run pre-training

   python H2SeqRec-Pretrain/run_pretrain.py
   b. Copy the output pre-training embedding to HGCN/data/AMT/use_pretrain/ folder

2. Hierarchical Hyperbolic Hypergraph Neural Network
   a. Prepare dataset.
   Split the dataset into monthly, quarterly and yearly dataset. For example, if the dataset contains 12 months, the HGCN/data/month folder will have 12 csv files, the HGCN/data/quarter folder will have 4 csv files and the HGCN/data/year folder will have 1 csv file. The format of each line is as follows. node1_id node2_id. The nodes within each hyperedge are considered connected with each other. The HGCN/config.py file is the configuration file.
   b. Modify parameters.
   Taking the monthly hyperbolic hypergraph embedding learning as an example, in HGCN/train.py change the pre-training embedding name in line 195, and modify the time_list in line 200. Supposing there are 12 months, the time_list should be range(1,13).
   c. Run hyperbolic hypergraph neural network.
   python HGCN/train.py --dataset AMT
   P.S. Run monthly, quarterly, yearly hyperbolic hypergraph neural network seperately. The steps are the same, but parameters should be modified.

3. User-side Hyperbolic Hypergraph Neural Network
   a. Prepare user-side hypergraph dataset.
   python create_user_side_hypergraph.py
   b. Modify parameters as 2-b.
   c. Run hyperbolic hypergraph neural network as 2-c.
   P.S. Only monthly is needed.

4. Completed Model
   a. Prepare needed data.
   python user_side_hypergraphs_new_hyperbolic_simple.py
   b. Run recommendation.
   python run_total.py