yoony02 / RL-2022-SRRL

This is the code for "Improving diversity and precision in sequential recommendation via RL"

Sequential recommendation with RL

A framework for diverse and accurate recommendation for users in sequential recommendation via reinforcement learning techniques

Setups

Datasets

The dataset name must be specified in the "--dataset" argument

• Beauty
• last.fm

After downloaded the datasets, you can put them in the folder data/ like the following.

$ tree
.
├── beauty
│   ├── preprocess_beauty.py
│   ├── Beauty.txt
│   └── item2attributes.json
└── lastfm
    ├── preprocess_lastfm.py
    ├── LastFM.txt
    └── item2attributes.json

And you can preprocess each datasets by running,

python preprocess_{dataset_name}.py

Train and Test

python main.py --dataset lastfm --gpu_num 1

Reference

• SA2C, SNQN (WSDM '22) [paper] [code]
  
• Datasets with attributes [link]

Project presenstation

Presentation PPT (in Korean)
Presentation Video (in Korean)