Group-Buying Recommendation for Social E-Commerce

This is the official implementation of the paper Group-Buying Recommendation for Social E-Commerce (PDF) accepted by ICDE'2021.

Group-Buying Dataset

Group buying, as an emerging form of purchase in social e-commerce websites, such as Pinduoduo.com , has recently achieved great success. In this new business model, users, initiator, can launch a group and share products to their social networks, and when there are enough friends, participants, join it, the deal is clinched. Group-buying recommendation for social ecommerce, which recommends an item list when users want to launch a group, plays an important role in the group success ratio and sales.

The information about the dataset can be found in BeiBei/readme.txt.

Code

We separate model definition from the framework librecframework for easily understanding.

You can find the framework librecframework in https://github.com/Sweetnow/librecframework.

Both modules mentioned in requirements.txt and librecframework should be installed before running the code.

More details about our codes will be added soon.

Usage

1. Download both librecframework and this repo

git clone [email protected]:Sweetnow/librecframework.git
git clone [email protected]:Sweetnow/group-buying-recommendation.git

2. Install librecframework (Python >= 3.8)

cd librecframework/
bash install.sh

3. Install dgl

4. Download negative.zip from Release, unzip it and copy *.negative.txt to datasets/BeiBei/

wget https://github.com/Sweetnow/group-buying-recommendation/releases/download/v1.0/negative.zip
unzip negative.zip
cp negative/* ${PATH-TO-GROUP-BUYING-RECOMMENDATION}/datasets/BeiBei

PS: negative sampling file is used for testing. More details can be found in Datasets README

4. Set config/config.json and config/pretrain.json following Docs.

5. Run the following command to know the CLI and check python environment:

python3 GBGCN train -h
# or
# python3 GBGCN test -h

PS: If you set hyperparameters that support multi input to multi values, the framework will automatically do grid-search accroding to your input. That is, use the Cartesian product of the hyperparameters for training and testing. For example, set --lr 0.1 0.01 -L 1 2, the codes will train and test model with hyperparameters [(0.1, 1), (0.1, 2), (0.01, 1), (0.01, 2)].

Citation

If you want to use our codes or dataset in your research, please cite:

@inproceedings{zhang2021group,
  title={Group-Buying Recommendation for Social E-Commerce},
  author={Zhang, Jun and Gao, Chen and Jin, Depeng and Li, Yong},
  booktitle={2021 IEEE 37th International Conference on Data Engineering (ICDE)},
  year={2021},
  organization={IEEE}
}

Acknowledgement