information_cocoon_study

Source code of the paper "An Exploratory Study on Information Cocoon in Recommender Systems".

1. running details

• config.py -- basic configuration: dataset, model, cuda, etc.

• rec_simulation.py -- interaction simulation between user and news recommender system.

2. train models

Before running 'rec_simulation.py', the recommendation model needs to be trained in advance.

• train_prob_predict.py -- train news recommendation model: NRMS, NAML, DKN.
• train_fm_and_ncf.py -- train DeepFM and NCF models.
• train_ngcf.py -- train NGCF model.

And the model implementation code is in 'model/' directory.

3. datasets

You can get the preprocessed data through Baidu Cloud Disk:

Link: https://pan.baidu.com/s/1tsW6CqFbG8OMYT1aTHhEQQ

Extraction code: kj7a

4. references

recommendation models refer to the following:

• NGCF: huangtinglin/NGCF-PyTorch: PyTorch Implementation for Neural Graph Collaborative Filtering (github.com)

• NCF and DeepFM: rixwew/pytorch-fm: Factorization Machine models in PyTorch (github.com)

• NRMS, NAML and DKN: yusanshi/news-recommendation: Implementations of some methods in news recommendation. (github.com)

• DPP: laming-chen/fast-map-dpp: Fast Greedy MAP Inference for DPP. (github.com)

• DGRec: YangLiangwei/DGRec (github.com)