AI帮你理解科学

AI 生成解读视频

AI抽取解析论文重点内容自动生成视频


pub
生成解读视频

AI 溯源

AI解析本论文相关学术脉络


Master Reading Tree
生成 溯源树

AI 精读

AI抽取本论文的概要总结


微博一下
We study the task of bundle recommender systems

Bundle Recommendation with Graph Convolutional Networks

international acm sigir conference on research and development in information retrieval, (2020)

被引用0|浏览328
下载 PDF 全文
引用
微博一下

摘要

Bundle recommendation aims to recommend a bundle of items for a user to consume as a whole. Existing solutions integrate user-item interaction modeling into bundle recommendation by sharing model parameters or learning in a multi-task manner, which cannot explicitly model the affiliation between items and bundles, and fail to explore th...更多

代码

数据

0
简介
  • The prevalence of bundled items on e-commerce and content platforms makes bundle recommendation become an important task.
  • Users need to be satisfied with most items in the bundle, which means that there is a sparser interaction between the user and the bundle.
  • Most existing works for bundle recommendation [2, 6, 7] regard item and bundle recommendation as two separate tasks, and associate them by sharing model parameters.
  • A recent study [3] proposed a multi-task framework that transfers the benefits of the item recommendation task to the bundle recommendation to alleviate the scarcity of user-bundle interactions.
  • The authors argue that they suffer from three major limitations:
重点内容
  • The prevalence of bundled items on e-commerce and content platforms makes bundle recommendation become an important task
  • A recent study [3] proposed a multi-task framework that transfers the benefits of the item recommendation task to the bundle recommendation to alleviate the scarcity of user-bundle interactions
  • We can observe that our model BGCN significantly outperforms all baselines in terms of Recall and NDCG metrics
  • We study the task of bundle recommender systems
  • We propose a graph-based solution BGCN that re-construct the two kinds of interaction and an affiliation into the graph
方法
  • MFBPR GCN-BG GCN-TG NGCF-BG NGCF-TG DAM BGCN % Improv. Netease

    Recall@20 NDCG@20 Recall@40 NDCG@40 Recall@80 NDCG@80 Youshu Model

    Recall@40 NDCG@40 Recall@40 NDCG@40

    Levels Propagation

    Item Level Bundle Level I&B Levels B2B Propagation

    No B2B Unweighted B2B Weighted B2B.
  • MFBPR GCN-BG GCN-TG NGCF-BG NGCF-TG DAM BGCN % Improv.
  • Recall@20 NDCG@20 Recall@40 NDCG@40 Recall@80 NDCG@80 Youshu Model.
  • Item Level Bundle Level I&B Levels B2B Propagation.
  • No B2B Unweighted B2B Weighted B2B
结果
  • Extensive experiments on two real-world datasets show that the proposed method outperforms existing state-of-the-art baselines by 10.77% to 23.18%.
  • The authors can observe that the model BGCN significantly outperforms all baselines in terms of Recall and NDCG metrics
结论
  • Conclusions and Future Work

    In this work, the authors study the task of bundle recommender systems.
  • The authors plan to consider the discount factor in bundle recommendation
总结
  • Introduction:

    The prevalence of bundled items on e-commerce and content platforms makes bundle recommendation become an important task.
  • Users need to be satisfied with most items in the bundle, which means that there is a sparser interaction between the user and the bundle.
  • Most existing works for bundle recommendation [2, 6, 7] regard item and bundle recommendation as two separate tasks, and associate them by sharing model parameters.
  • A recent study [3] proposed a multi-task framework that transfers the benefits of the item recommendation task to the bundle recommendation to alleviate the scarcity of user-bundle interactions.
  • The authors argue that they suffer from three major limitations:
  • Methods:

    MFBPR GCN-BG GCN-TG NGCF-BG NGCF-TG DAM BGCN % Improv. Netease

    Recall@20 NDCG@20 Recall@40 NDCG@40 Recall@80 NDCG@80 Youshu Model

    Recall@40 NDCG@40 Recall@40 NDCG@40

    Levels Propagation

    Item Level Bundle Level I&B Levels B2B Propagation

    No B2B Unweighted B2B Weighted B2B.
  • MFBPR GCN-BG GCN-TG NGCF-BG NGCF-TG DAM BGCN % Improv.
  • Recall@20 NDCG@20 Recall@40 NDCG@40 Recall@80 NDCG@80 Youshu Model.
  • Item Level Bundle Level I&B Levels B2B Propagation.
  • No B2B Unweighted B2B Weighted B2B
  • Results:

    Extensive experiments on two real-world datasets show that the proposed method outperforms existing state-of-the-art baselines by 10.77% to 23.18%.
  • The authors can observe that the model BGCN significantly outperforms all baselines in terms of Recall and NDCG metrics
  • Conclusion:

    Conclusions and Future Work

    In this work, the authors study the task of bundle recommender systems.
  • The authors plan to consider the discount factor in bundle recommendation
表格
  • Table1: Statistics of two utilized real-world datasets
  • Table2: Performance comparisons on two real-world datasets with six baselines
  • Table3: Ablation study of the key designs
Download tables as Excel
相关工作
  • Although bundles are currently widely used everywhere, few efforts have been made in solving the bundle recommendation problem. List Recommendation Model (LIRE) [6] and Embedding Factorization Machine (EFM) [2] simultaneously utilized the users’ interactions with both items and bundles under the BPR framework. The Bundle BPR (BBPR) Model [7] made use of the parameters

    Average Recall@40 Average NDCG@40

    MFBPR GCN-BG NGCF-BG GCN-TG NGCF-TG DAM BGCN 3~5 >5

    #Records of Bundles 0~3

    previously learned through an item BPR model. Recently, Deep

    Attentive Multi-Task Model (DAM) [3] jointly modeled user-bundle interactions and user-item interactions in a multi-task manner.

    The basic idea of graph convolutional networks (GCN) [5] is to reduce the high-dimensional adjacency information of a node in the graph to a low-dimensional vector representation. With the strong power of learning structure, GCN is widely applied in recommender systems. Berg et al [1] first applied GCN to the recommendation to factorize several rating matrices. Ying et al [10] extended it to web-scale recommender systems with neighbor-sampling. Recently, Wang et al [9] further approached a more general model that uses high-level connectivity learned by GCN to encode CF signals.
引用论文
  • Rianne van den Berg, Thomas N Kipf, and Max Welling. 2017. Graph convolutional matrix completion. In KDD.
    Google ScholarFindings
  • Da Cao, Liqiang Nie, Xiangnan He, Xiaochi Wei, Shunzhi Zhu, and Tat-Seng Chua. 2017. Embedding factorization models for jointly recommending items and user generated lists. In SIGIR. 585–594.
    Google ScholarLocate open access versionFindings
  • Liang Chen, Yang Liu, Xiangnan He, Lianli Gao, and Zibin Zheng. 2019. Matching user with item set: collaborative bundle recommendation with deep attention network. In IJCAI. 2095–2101.
    Google ScholarFindings
  • Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, and Tat-Seng Chua. 2017. Neural collaborative filtering. In WWW. 173–182.
    Google ScholarLocate open access versionFindings
  • Thomas N Kipf and Max Welling. 2017. Semi-supervised classification with graph convolutional networks. In ICLR.
    Google ScholarFindings
  • Yidan Liu, Min Xie, and Laks VS Lakshmanan. 2014. Recommending user generated item lists. In RecSys. 185–192.
    Google ScholarLocate open access versionFindings
  • Apurva Pathak, Kshitiz Gupta, and Julian McAuley. 201Generating and personalizing bundle recommendations on steam. In SIGIR. 1073–1076.
    Google ScholarFindings
  • Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme. 2009. BPR: Bayesian personalized ranking from implicit feedback. In UAI.
    Google ScholarFindings
  • Xiang Wang, Xiangnan He, Meng Wang, Fuli Feng, and Tat-Seng Chua. 201Neural graph collaborative filtering. In SIGIR. 165–174.
    Google ScholarLocate open access versionFindings
  • Rex Ying, Ruining He, Kaifeng Chen, Pong Eksombatchai, William L Hamilton, and Jure Leskovec. 2018. Graph convolutional neural networks for web-scale recommender systems. In KDD. 974–983.
    Google ScholarLocate open access versionFindings
您的评分 :
0

 

标签
评论
数据免责声明
页面数据均来自互联网公开来源、合作出版商和通过AI技术自动分析结果,我们不对页面数据的有效性、准确性、正确性、可靠性、完整性和及时性做出任何承诺和保证。若有疑问,可以通过电子邮件方式联系我们:report@aminer.cn
小科