DIVAN: Deep-Interest Virality-Aware Network to Exploit Temporal Dynamics in News Recommendation

This is the official repository for the paper "DIVAN: Deep-Interest Virality-Aware Network to Exploit Temporal Dynamics in News Recommendation", published at the ACM RecSys Challenge 2024 (RecSys Challenge ’24).

DIVAN (Deep-Interest Virality-Aware Network), our solution for the RecSys 2024 Challenge, combines a Deep Interest Network (DIN) for personalized user interest representation with a Virality-Aware Click Predictor that utilizes temporal features to estimate click probability based on news popularity. A user-specific weight balances the influence of DIN and virality-based predictions, enhancing personalization and accuracy. Experiments on the Ekstra Bladet dataset from the Challenge demonstrate how promising DIVAN is in accuracy and beyond-accuracy performance.

Find more details in the paper: https://dl.acm.org/doi/10.1145/3687151.3687153.

This repository is built on top of FuxiCTR, a configurable, tunable, and reproducible library for CTR prediction.

Paper reference FuxiCTR:

• Jieming Zhu, Jinyang Liu, Shuai Yang, Qi Zhang, Xiuqiang He. Open Benchmarking for Click-Through Rate Prediction. The 30th ACM International Conference on Information and Knowledge Management (CIKM), 2021.

If you use any part of this code, please cite the work:

@inproceedings{10.1145/3687151.3687153,
author = {Ferrara, Antonio and Valentini, Marco and Masciullo, Paolo and De Candia, Antonio and Abbattista, Davide and Fusco, Riccardo and Pomo, Claudio and Anelli, Vito Walter and Biancofiore, Giovanni Maria and Boratto, Ludovico and Narducci, Fedelucio},
title = {DIVAN: Deep-Interest Virality-Aware Network to Exploit Temporal Dynamics in News Recommendation},
year = {2024},
isbn = {9798400711275},
publisher = {Association for Computing Machinery},
url = {https://doi.org/10.1145/3687151.3687153},
doi = {10.1145/3687151.3687153},
booktitle = {Proceedings of the Recommender Systems Challenge 2024},
pages = {12–16},
series = {RecSysChallenge '24}
}

Setup virtual environment

If you want to use venv

1. Please set up the environment as follows (we used python 3.9 and python 3.10).

   python3 -m venv recsys_din
   source recsys_din/bin/activate
   python -m pip install --upgrade pip
   pip install --no-cache-dir -r requirements.txt
   

If you want to use Docker

1. make sure you have started the docker engine
2. Build the container:

      docker build -t recsyschallenge2024_din .

3. Run the container

      docker run -d -it --name recsyschallenge_container  recsyschallenge2024_din /bin/bash

4. Access the terminal of the container

   docker exec -it recsyschallenge_container /bin/sh

Data Preparation and Model Training

1. Download and preprare data

   python prepare_data_v1.py --size large --test --embedding_size 64 --neg_sampling

2. Train the model on train and validation sets:

3. python run_param_tuner.py --config config/DIVAN_ebnerd_large_x1_tuner_config_01.yaml --gpu 0

4. Make predictions on the test set:

   Get the experiment_id from running logs or the result csv file DIVAN_ebnerd_large_x1_tuner_config_01.csv, and then you can run prediction on the test.

   python submit.py --config config/DIVAN_ebnerd_large_x1_tuner_config_01 --expid DIVAN_ebnerd_large_x1_001_1860e41e --gpu 0

Data preparation and prediction with PopularRanker and ViralRanker

1. Download and preprare data

   python prepare_data_pop_and_vir_scores.py --size large --test

2. Test the model on the validation set:

   python run_[popular|virality]_expid.py

3. Make predictions on the test set:

    python submit_[popular|viral].py