DRGR: Deep Reinforcement learning based Group Recommender system

A course project for Georgia Tech CSE 6240 Web Search and Text Mining Spring 2021.

Dataset

The original MovieLens dataset can be downloaded from MovieLens, and the one we use is MovieLens 1M Dataset. You can generate the group data MovieLens-Rand by using data/MovieLens-1M.zip and drgr/generator.py. (See "Run the DRGR".)

The data/MovieLens-Rand directory should include the following files:

movies.dat:

• Movie information file from MovieLens-1M.

users.dat:

• User information file from MovieLens-1M.

groupMember.dat:

• File including group members.
• Each line is a group instance: groupID userID1,userID2,...

group(user)RatingTrain.dat:

• Train file.
• Each line is a training instance: groupID(userID) itemID rating timestamp

group(user)RatingVal(Test).dat:

• group (user) validation (test) file (positive instances).
• Each line is a validation (test) instance: groupID(userID) itemID rating timestamp

group(user)RatingVal(Test)Negative.dat:

• group (user) validation (test) file (negative instances).
• Each line corresponds to the line of group(user)RatingVal(Test).dat, containing 100 negative samples.
• Each line is in the format: (groupID(userID),itemID) negativeItemID1 negativeItemID2 ...

Run the DRGR

Change to the code directory (if not in):

cd drgr

Generate the group data from the MovieLens-1M dataset (data/MovieLens-1M.zip):

python generator.py

Run the DRGR:

python main.py

Parameters

We put all parameters in config.py.

Tree

.
├── [5.6M]  images
│   ├── [5.5M]  deps.png
│   ├── [ 30K]  group_recommender_actorcritic_1.svg
│   ├── [ 10K]  group_recommender_actorcritic_2.svg
│   ├── [8.1K]  group_recommender_actorcritic_3.svg
│   ├── [ 12K]  group_recommender_actorcritic_4.svg
│   ├── [ 14K]  group_recommender_actorcritic_5.svg
│   ├── [ 13K]  group_recommender_actorcritic_6.svg
│   ├── [ 12K]  group_recommender_actorcritic_7.svg
│   ├── [ 21K]  group_recommender_actorcritic_8.svg
│   └── [5.9K]  S758139 | RL-based Group Recommender.svg
├── [150K]  nbs
│   └── [146K]  T381683_Group_Recommendations_with_Actor_critic_RL_Agent_in_MDP_Environment_on_ML_1m_Dataset.ipynb
├── [2.2K]  README.md
└── [ 52K]  src
    ├── [8.7K]  agent.py
    ├── [2.1K]  config.py
    ├── [ 10K]  data.py
    ├── [4.0K]  env.py
    ├── [2.6K]  eval.py
    ├── [ 13K]  generator.py
    ├── [2.3K]  main.py
    ├── [3.3K]  model.py
    └── [1.9K]  utils.py

 5.8M used in 3 directories, 21 files

References

During the code implementation, we reference the following repositories and articles:

1. https://arxiv.org/abs/2106.06900)
2. AGREE: https://github.com/LianHaiMiao/Attentive-Group-Recommendation
3. LIRD: https://github.com/egipcy/LIRD
4. Recsys-RL: https://github.com/shashist/recsys-rl
5. GroupIM: https://github.com/CrowdDynamicsLab/GroupIM
6. Deep Deterministic Policy Gradients Explained: https://towardsdatascience.com/deep-deterministic-policy-gradients-explained-2d94655a9b7b