UISA: User Information Separating Architecture for Commodity Recommendation Policy with Deep Reinforcement Learning
ACM Transactions on Recommender Systems, 2024. PDF
python=3.6
gym=0.17.3
torch=1.8.2
wandb=0.12.2
pandas=1.1.1
numpy=1.19.2
The environment is based on JData. The e-commerce website is https://www.jd.com/, and the JData dataset can be found at https://www.kaggle.com/datasets/owincontext/jdata2016.
The files in the directory /JDataExp/data/ are part of environemnt.
Models are in the directory /JDataExp/data/Newmodel.
/JDataExp/model/DataProcessing.py and /JDataExp/model/DataProcessing.py are the codes for construct JDEnv. If you want to construct JDEnv from scratch, please download JData and put it in /JDataExp/data.
The environment cites the work of "Jing-Cheng Shi, Yang Yu, Qing Da, Shi-Yong Chen, and An-Xiang Zeng. Virtual-Taobao: Virtualizing real-world online retail environment for reinforcement learning. In: Proceedings of the 33rd AAAI Conference on Artificial Intelligence (AAAI’19), Honolulu, HI, 2019.", the environment model is same as https://github.com/eyounx/VirtualTaobao.
Our models are in the directory /virtualTB/Newmodel.
/virtualTB/Newmodel/GetData.py generate the testdata.