ytchx1999 / CLSR

Implementation of "Disentangling Long and Short-Term Interests for Recommendation" (WWW '22)

CLSR: Disentangling Long and Short-Term Interests for Recommendation

This is the official implementation of our WWW'22 paper:

Yu Zheng, Chen Gao, Jianxin Chang, Yanan Niu, Yang Song, Depeng Jin, Yong Li, Disentangling Long and Short-Term Interests for Recommendation, In Proceedings of the Web Conference 2022.

The code is tested under a Linux desktop with TensorFlow 1.15.2 and Python 3.6.8.

环境

tensorflow-gpu==1.15.0
pandas==1.1.5
PyYAML==6.0
requests==2.27.1
scikit-learn==0.20.4
tqdm==4.64.0

pip install -r requirements.txt

Data Pre-processing

Run the script reco_utils/dataset/sequential_reviews.py to generate the data for training and evaluation.

新建目录tests/resources/deeprec/sequential/taobao和tests/resources/deeprec/sequential/kuaishou

taobao数据集：https://tianchi.aliyun.com/dataset/dataDetail?dataId=649

Model Training

Use the following command to train a CLSR model on Taobao dataset:

# python examples/00_quick_start/sequential.py --dataset taobao

cd examples/00_quick_start/
python3 sequential.py --dataset taobao

or on Kuaishou dataset:

python examples/00_quick_start/sequential.py --dataset kuaishou

Note

The implemention is based on Microsoft Recommender.