Paper Recommendation

Given a user-paper bipartite graph G = {U, V, R} with the sets of users U, papers V, and the interactions R (i.e, click). Given R, the objective of recommender systems is to predict the possible papers that the user will interact next.

We collect scientific resource reading behavior from AMiner of its users from August to October 2021. To ensure the quality, the dataset uses the 10-core setting, i.e., it retains users and items with at least ten interactions. Furthermore, to unify the experimental settings, it splits the interacted records as training, test, and validation set and ensures that 10% of interactions as the test set for each user to evaluate the recommendation performance, and another 10% of interactions are used as a validation set, and the remaining 80% interactions are used as the training set.

Statistics:

users: 5,340

items: 14,967

interactions:163,084

URL: https://www.aminer.cn/data/?nav=openData#AMiner-Paper-Click

We apply two widely-used metrics Recall@K and NDCG@K to evaluate the top-K performance.

Mult-VAE: is an item-based CF method for implicit feedback by variational autoencoders (VAEs). It adopts the multi-nomial likelihood of data, uses Bayesian inference for parameter estimation, and connects information-theoretic to maximum entropy discrimination.

GF-CF: proposes graph filter-based CF and proves the special cases via the lens of graph signal processing as well as the importance of smoothness.

NGCF: proposes a message-passing architecture to capture collaborative filtering signal in the first-order propagation and high-order propagation. NGCF stacks embeddings of multiple propagation layers as its final representation for users and items.

LightGCN: simplifies the design GCN that abandons feature transformation and nonlinear activation, adopts the most essential component neighborhood aggregation for collaborative filtering, and uses a weighted sum of embeddings of all layers as its final embeddings.

Code & Data: https://github.com/THUDM/Paper-Rec

Data: https://lfs.aminer.cn//misc/AMiner-Paper-Click.zip    

We kindly request that any published research that makes use of this data cites the following papers.  

@inproceedings{zhang2024oag,
  title={OAG-bench: a human-curated benchmark for academic graph mining},
  author={Fanjin Zhang and Shijie Shi and Yifan Zhu and Bo Chen and Yukuo Cen and Jifan Yu and Yelin Chen and Lulu Wang and Qingfei Zhao and Yuqing Cheng and Tianyi Han and Yuwei An and Dan Zhang and Weng Lam Tam and Kun Cao and Yunhe Pang and Xinyu Guan and Huihui Yuan and Jian Song and Xiaoyan Li and Yuxiao Dong and Jie Tang},
  booktitle={Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
  pages={6214--6225},
  year={2024}
}

@inproceedings{zhang2023apegnn,
  title={ApeGNN: node-wise adaptive aggregation in GNNs for recommendation},
  author={Zhang, Dan and Zhu, Yifan and Dong, Yuxiao and Wang, Yuandong and Feng, Wenzheng and Kharlamov, Evgeny and Tang, Jie},
  booktitle={Proceedings of the ACM Web Conference 2023},
  pages={759--769},
  year={2023}
}