recommendation_dataset_for_pre-training & transfer learning & lifelong learning & cross-domain recommendation & cold-start recommendation
Some New datasets, go to https://github.com/westlake-repl
DataSets links for recommender systems research, in particular for transfer learning, user representation, pre-training,lifelong learning, cold start recommendation
https://drive.google.com/file/d/1imhHUsivh6oMEtEW-RwVc4OsDqn-xOaP/view?usp=sharin
A large-scale recommmendation datasets used in
@inproceedings{yuan2021one,
title={One person, one model, one world: Learning continual user representation without forgetting},
author={Yuan, Fajie and Zhang, Guoxiao and Karatzoglou, Alexandros and Jose, Joemon and Kong, Beibei and Li, Yudong},
booktitle={Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval},
pages={696--705},
year={2021}
}
@inproceedings{yuan2020parameter,
title={Parameter-efficient transfer from sequential behaviors for user modeling and recommendation},
author={Yuan, Fajie and He, Xiangnan and Karatzoglou, Alexandros and Zhang, Liguang},
booktitle={Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval},
pages={1469--1478},
year={2020}
}
including over 1 million users and around 0.5million items and almost 100 million behaviors.
The dataset is of high quality and you may find that by pretraining in advance could really benefit the new recommendation tasks. It can be used to evaluate deep learning models built for transfer learning, pretraining, cross-domain recommendation, multi-task recommendation and lifelong learning, cold-user recommendation and user profile prediction, session-based recommendation etc.