LSEC-GNN

The repository contains the details about our paper Leveraging Tripartite Interaction Information from Live Stream E-Commerce for Improving Product Recommendation (KDD 2021).

Requirements

• Linux-based OS
• Python 3.6+

Get started

Clone the project

git clone https://github.com/yusanshi/LSEC-GNN && cd LSEC-GNN

Install the package

Note this repository contains no model code. Instead we organized our code into a separate Python package called RecHub. It is a general package for recommendation featuring GNN training especially heterogeneous graph. Refer to its repository for the details.

# Installing the main package.
pip install rechub

# Installing the DGL library.
# This is for CPU version. For CUDA version, use dgl-cu[xxx]
pip install dgl # or dgl-cu92, dgl-cu101, dgl-cu102, dgl-cu110 for CUDA 9.2, 10.1, 10.2, 11.0, respectively.

If there are any problems with later commands, try to install this specific version (PyPI, GitHub):

pip install rechub==0.0.5
pip install dgl==0.5.3 # or CUDA version: dgl-cu[xxx]==0.5.3

Download the dataset

mkdir data && cd data
wget https://github.com/yusanshi/LSEC-GNN/files/6520753/LSEC-Small-aa.dummy.gz \
 https://github.com/yusanshi/LSEC-GNN/files/6520754/LSEC-Small-ab.dummy.gz \
 https://github.com/yusanshi/LSEC-GNN/files/6520757/LSEC-Small-ac.dummy.gz \
 https://github.com/yusanshi/LSEC-GNN/files/6520760/LSEC-Small-ad.dummy.gz
cat LSEC-Small-* | tar -xzvf -

Train

Each line of commands corresponds to a method in the above table (taken from the paper).

# In the project root directory
python -m rechub.train --model_name NCF --dataset_path ./data/LSEC-Small --metadata_path ./metadata/LSEC.json --edge_choice 0 --training_task_choice 0 --evaluation_task_choice 0
python -m rechub.train --model_name GCN --dataset_path ./data/LSEC-Small --metadata_path ./metadata/LSEC.json --edge_choice 0 --training_task_choice 0 --evaluation_task_choice 0
python -m rechub.train --model_name LightGCN --dataset_path ./data/LSEC-Small --metadata_path ./metadata/LSEC.json --edge_choice 0 --training_task_choice 0 --evaluation_task_choice 0
python -m rechub.train --model_name GAT --dataset_path ./data/LSEC-Small --metadata_path ./metadata/LSEC.json --edge_choice 0 --training_task_choice 0 --evaluation_task_choice 0
python -m rechub.train --model_name HET-GCN --dataset_path ./data/LSEC-Small --metadata_path ./metadata/LSEC.json --edge_choice 0 1 2 --training_task_choice 0 1 --evaluation_task_choice 0
python -m rechub.train --model_name HET-LightGCN --dataset_path ./data/LSEC-Small --metadata_path ./metadata/LSEC.json --edge_choice 0 1 2 --training_task_choice 0 1 --evaluation_task_choice 0
python -m rechub.train --model_name HET-GAT --dataset_path ./data/LSEC-Small --metadata_path ./metadata/LSEC.json --edge_choice 0 1 2 --training_task_choice 0 1 --evaluation_task_choice 0

You can visualize metrics with TensorBoard.

tensorboard --logdir runs

Tip: by adding REMARK environment variable, you can make the runs name in TensorBoard more meaningful. For example, REMARK=lr-0.001 python -m rechub.train ....

Test

You can run the training commands with parameter --save_checkpoint True and use the saved checkpoints to evaluate after training.

Here we provide the checkpoints so you can directly download and evaluate with them.

First download checkpoint.tar.gz from https://drive.google.com/file/d/1RA-wF1hQxLkjVN7R3N1xHkhPDjsEqVv8/view?usp=sharing. Put it into project root directory.

Now extract the checkpoints:

# In the project root directory
mkdir checkpoint
tar -xzvf checkpoint.tar.gz -C checkpoint

Then replace rechub.train in the training commands with rechub.test to evaluate the checkpoints.

Note the results you get with the checkpoints could have some difference with the results in paper, since the latter are averaged results.

Credits

JD Technology (京东科技)

Cite

@inproceedings{LSEC-GNN,
author = {Yu, Sanshi and Jiang, Zhuoxuan and Chen, Dong-Dong and Feng, Shanshan and Li, Dongsheng and Liu, Qi and Yi, Jinfeng},
title = {Leveraging Tripartite Interaction Information from Live Stream E-Commerce for Improving Product Recommendation},
year = {2021},
isbn = {9781450383325},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3447548.3467151},
doi = {10.1145/3447548.3467151},
booktitle = {Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining},
pages = {3886–3894},
numpages = {9},
keywords = {multi-task learning, live streaming e-commence, graph representation learning, product recommendation},
location = {Virtual Event, Singapore},
series = {KDD '21}
}