PNS4Rec:利用Personalized Next Session改善基於Session的推薦系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	施柏丞 Po-Cheng Shih
論文名稱：	PNS4Rec:利用Personalized Next Session改善基於Session的推薦系統 PNS4Rec: Personalized Next Session for Session-based Recommendation
指導教授：	陳彥良 Yen-Liang Chen
口試委員:
學位類別：	碩士 Master
系所名稱：	管理學院 - 資訊管理學系 Department of Information Management
論文出版年：	2022
畢業學年度：	110
語文別：	中文
論文頁數：	55
中文關鍵詞：	基於Session的推薦、深度學習、個性化推薦、神經網路、全局資訊、Session 嵌入
外文關鍵詞：	Session-based recommendation, Deep learning, Personalized recommendation, Neural network, Global information, Session embedding
相關次數：	點閱：15 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

在電子商務中，由於Session中有關用戶的資訊有限，要提供基於Session的個人化推薦仍然是一項具有挑戰性的工作。因此，本文認為僅僅基於用戶個人的局部資訊（使用者個人行為），即只包含個人的長期資訊以及短期資訊是遠遠不夠的，應該納入全局資訊作為考慮，進而從現有的Session中，挖掘出更多可利用的協作訊息。
然而我們並非參考所有的全局資訊，而是從中挑選對用戶有用的Session，即所謂的Personalized Next Session (PNS)做為全局資訊，以改善局部資訊的不足。
本文是第一個採用包含PNS概念的深度網路架構研究，並同時結合局部資訊與全局資訊，以向使用者推薦當前Session的下一個可能交互項目。我們對多個真實世界的資料集進行了實驗，大量的實驗結果表明我們提出的深度網路架構性能優於幾種最先進的推薦方法。

In e-commerce, personalized session-based recommendation is still a challenging task because of the limited information about the user in the session. Therefore, in this study, we believe that it is far from enough to include only the user's local information (user's personal behavior), i.e., only personal long-term information and short-term information, and we should take global information into consideration to explore more collaborative information from existing sessions. However, we do not refer to all the global information, but select sessions that are useful to users, the so-called Personalized Next Session (PNS) as the global information, in order to improve the shortage of local information. This work is the first to adopt a deep network architecture research that includes the concept of PNS, and combines local and global information at the same time to recommend the next item of the current session to the user. We have conducted experiments on several real-world datasets. Extensive experimental results show that our proposed deep network architecture outperforms several of the most advanced recommendation methods.

Contents
摘要    i
List of Figures    v
List of Tables    vi
1.    Introduction    1
2.    Related work    7
2-1 Traditional methods    7
2-2 Neural network methods    8
2-2-1 Recurrent neural network based methods    8
2-2-2 Graph neural network based methods    10
3.    Proposed approach    12
3-1 Overview    12
3-2 Preprocessing    17
3-2-1 Graph embedding    17
3-2-2 Neighborhood Session (NS)    18
3-2-3 Personalized Next Session (PNS)    19
3-3 Model    19
3-3-1 Local Information Module (LIM)    19
3-3-2 Personalized Next Session Module (PNSM)    21
3-3-3 Preference prediction    22
3-3-4 Network training    23
4.    Experiments    24
4-1 Datasets    24
4-2 Research questions and evaluation metrics    25
4-3 Baselines and parameter settings    27
4-4 Experimental platform    28
5.    Results    29
5-1 Ablation study    29
5-2 Performance of models    31
5-3 Sensitivity analysis    33
5-3-1 Impact of Neighborhood Session number j    33
5-3-2 Impact of PNS item number k    34
5-3-3 Impact of history number L    35
5-3-4 Impact of embedding size D and batch size B    35
6.    Conclusion and future work    38
6-1 Conclusion    38
6-2 Future work    39
Reference    40


                                

Reference
[1] M. H. Mohamed, M. H. Khafagy及M. H. Ibrahim, 作者, 「Recommender Systems Challenges and Solutions Survey」, 收入 2019 International Conference on Innovative Trends in Computer Engineering (ITCE), Aswan, Egypt, 2月 2019, 頁 149–155. doi: 10.1109/ITCE.2019.8646645.
[2] J. Li, W. Xu, W. Wan及J. Sun, 作者, 「Movie recommendation based on bridging movie feature and user interest」, Journal of Computational Science, 卷 26, 頁 128–134, 5月 2018, doi: 10.1016/j.jocs.2018.03.009.
[3] H.-T. Zheng, J.-Y. Chen, N. Liang, A. Sangaiah, Y. Jiang及C.-Z. Zhao, 作者, 「A Deep Temporal Neural Music Recommendation Model Utilizing Music and User Metadata」, Applied Sciences, 卷 9, 期 4, 頁 703, 2月 2019, doi: 10.3390/app9040703.
[4] M. Karimi, D. Jannach及M. Jugovac, 作者, 「News recommender systems – Survey and roads ahead」, Information Processing & Management, 卷 54, 期 6, 頁 1203–1227, 11月 2018, doi: 10.1016/j.ipm.2018.04.008.
[5] J. Fan, Y. Jiang, Y. Liu及Y. Zhou, 作者, 「Interpretable MOOC recommendation: a multi-attention network for personalized learning behavior analysis」, INTR, 卷 ahead-of-print, 期 ahead-of-print, 6月 2021, doi: 10.1108/INTR-08-2020-0477.
[6] S. Rendle, C. Freudenthaler, Z. Gantner及L. Schmidt-Thieme, 作者, 「BPR: Bayesian Personalized Ranking from Implicit Feedback」, 頁 10, 2009.
[7] B. Sarwar, G. Karypis, J. Konstan及J. Reidl, 作者, 「Item-based collaborative filtering recommendation algorithms」, 收入 Proceedings of the tenth international conference on World Wide Web - WWW ’01, Hong Kong, Hong Kong, 2001, 頁 285–295. doi: 10.1145/371920.372071.
[8] M. Ludewig及D. Jannach, 作者, 「Evaluation of session-based recommendation algorithms」, User Model User-Adap Inter, 卷 28, 期 4–5, 頁 331–390, 12月 2018, doi: 10.1007/s11257-018-9209-6.
[9] S. Wang, L. Cao, Y. Wang, Q. Z. Sheng, M. A. Orgun及D. Lian, 作者, 「A Survey on Session-based Recommender Systems」, ACM Comput. Surv., 卷 54, 期 7, 頁 1–38, 9月 2022, doi: 10.1145/3465401.

[10] B. Hidasi, A. Karatzoglou, L. Baltrunas及D. Tikk, 作者, 「Session-based Recommendations with Recurrent Neural Networks」, arXiv:1511.06939 [cs], 3月 2016, 引見於: 2021年10月4日. [線上]. 載於: http://arxiv.org/abs/1511.06939
[11] J. Li, P. Ren, Z. Chen, Z. Ren, T. Lian及J. Ma, 作者, 「Neural Attentive Session-based Recommendation」, 收入 Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, Singapore Singapore, 11月 2017, 頁 1419–1428. doi: 10.1145/3132847.3132926.
[12] Y. K. Tan, X. Xu及Y. Liu, 作者, 「Improved Recurrent Neural Networks for Session-based Recommendations」, 收入 Proceedings of the 1st Workshop on Deep Learning for Recommender Systems, Boston MA USA, 9月 2016, 頁 17–22. doi: 10.1145/2988450.2988452.
[13] Q. Liu, Y. Zeng, R. Mokhosi及H. Zhang, 作者, 「STAMP: Short-Term Attention/Memory Priority Model for Session-based Recommendation」, 收入 Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, London United Kingdom, 7月 2018, 頁 1831–1839. doi: 10.1145/3219819.3219950.
[14] M. Hosseinzadeh Aghdam, N. Hariri, B. Mobasher及R. Burke, 作者, 「Adapting Recommendations to Contextual Changes Using Hierarchical Hidden Markov Models」, 收入 Proceedings of the 9th ACM Conference on Recommender Systems, Vienna Austria, 9月 2015, 頁 241–244. doi: 10.1145/2792838.2799684.
[15] S. Rendle, C. Freudenthaler及L. Schmidt-Thieme, 作者, 「Factorizing personalized Markov chains for next-basket recommendation」, 收入 Proceedings of the 19th international conference on World wide web - WWW ’10, Raleigh, North Carolina, USA, 2010, 頁 811. doi: 10.1145/1772690.1772773.
[16] D. Jannach, L. Lerche及M. Jugovac, 作者, 「Adaptation and Evaluation of Recommendations for Short-term Shopping Goals」, 收入 Proceedings of the 9th ACM Conference on Recommender Systems, Vienna Austria, 9月 2015, 頁 211–218. doi: 10.1145/2792838.2800176.
[17] Y. Song, A. M. Elkahky及X. He, 作者, 「Multi-Rate Deep Learning for Temporal Recommendation」, 收入 Proceedings of the 39th International ACM SIGIR conference on Research and Development in Information Retrieval, Pisa Italy, 7月 2016, 頁 909–912. doi: 10.1145/2911451.2914726.

[18] J. Song, H. Shen, Z. Ou, J. Zhang, T. Xiao及S. Liang, 作者, ISLF: Interest Shift and Latent Factors Combination Model for Session-based Recommendation. 2019, 頁 5771. doi: 10.24963/ijcai.2019/799.
[19] D. Hu, L. Wei, W. Zhou, X. Huai, Z. Fang及S. Hu, 作者, 「PEN4Rec: Preference Evolution Networks for Session-based Recommendation」, arXiv:2106.09306 [cs], 6月 2021, 引見於: 2021年11月21日. [線上]. 載於: http://arxiv.org/abs/2106.09306
[20] S. Feng, X. Li, Y. Zeng, G. Cong, Y. M. Chee及Q. Yuan, 作者, 「Personalized Ranking Metric Embedding for Next New POI Recommendation」, 頁 7.
[21] D. Jannach及M. Ludewig, 作者, 「When Recurrent Neural Networks meet the Neighborhood for Session-Based Recommendation」, 收入 Proceedings of the Eleventh ACM Conference on Recommender Systems, Como Italy, 8月 2017, 頁 306–310. doi: 10.1145/3109859.3109872.
[22] M. Ye, X. Liu及W.-C. Lee, 作者, 「Exploring social influence for recommendation: a generative model approach」, 頁 10.
[23] W. Song, Z. Xiao, Y. Wang, L. Charlin, M. Zhang及J. Tang, 作者, 「Session-based Social Recommendation via Dynamic Graph Attention Networks」, 頁 9, 2019.
[24] Z. Wang等, 作者, 「Exploring Global Information for Session-based Recommendation」, arXiv:2011.10173 [cs], 6月 2021, 引見於: 2021年11月21日. [線上]. 載於: http://arxiv.org/abs/2011.10173
[25] Y. Wu及J. Gou, 作者, 「Leveraging neighborhood session information with dual attentive neural network for session-based recommendation」, Neurocomputing, 卷 439, 頁 234–242, 6月 2021, doi: 10.1016/j.neucom.2021.01.051.
[26] G. Shani, D. Heckerman及R. I. Brafman, 作者, 「An MDP-Based Recommender System」, 頁 31.
[27] F. Eskandanian及B. Mobasher, 作者, 「Detecting Changes in User Preferences using Hidden Markov Models for Sequential Recommendation Tasks」, arXiv:1810.00272 [cs], 9月 2018, 引見於: 2021年11月14日. [線上]. 載於: http://arxiv.org/abs/1810.00272
[28] R. Chen, Q. Hua, Y.-S. Chang, B. Wang, L. Zhang及X. Kong, 作者, 「A Survey of Collaborative Filtering-Based Recommender Systems: From Traditional Methods to Hybrid Methods Based on Social Networks」, IEEE Access, 卷 6, 頁 64301–64320, 2018, doi: 10.1109/ACCESS.2018.2877208.
[29] M. Wang, P. Ren, L. Mei, Z. Chen, J. Ma及M. de Rijke, 作者, 「A Collaborative Session-based Recommendation Approach with Parallel Memory Modules」, 收入 Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, Paris France, 7月 2019, 頁 345–354. doi: 10.1145/3331184.3331210.
[30] M. Ruocco, O. S. L. Skrede及H. Langseth, 作者, 「Inter-Session Modeling for Session-Based Recommendation」, 收入 Proceedings of the 2nd Workshop on Deep Learning for Recommender Systems, Como Italy, 8月 2017, 頁 24–31. doi: 10.1145/3125486.3125491.
[31] Q. Cui, S. Wu, Y. Huang及L. Wang, 作者, 「A hierarchical contextual attention-based network for sequential recommendation」, Neurocomputing, 卷 358, 頁 141–149, 9月 2019, doi: 10.1016/j.neucom.2019.04.073.
[32] N. Bhatia及C. Author, 作者, 「Survey of Nearest Neighbor Techniques」, 卷 8, 期 2, 頁 4, 2010.
[33] M. Quadrana, A. Karatzoglou, B. Hidasi及P. Cremonesi, 作者, 「Personalizing Session-based Recommendations with Hierarchical Recurrent Neural Networks」, 收入 Proceedings of the Eleventh ACM Conference on Recommender Systems, Como Italy, 8月 2017, 頁 130–137. doi: 10.1145/3109859.3109896.
[34] Y. Guo, D. Zhang, Y. Ling及H. Chen, 作者, 「A Joint Neural Network for Session-Aware Recommendation」, IEEE Access, 卷 8, 頁 74205–74215, 2020, doi: 10.1109/ACCESS.2020.2984287.
[35] H. Cai, V. W. Zheng及K. C.-C. Chang, 作者, 「A Comprehensive Survey of Graph Embedding: Problems, Techniques, and Applications」, IEEE Trans. Knowl. Data Eng., 卷 30, 期 9, 頁 1616–1637, 9月 2018, doi: 10.1109/TKDE.2018.2807452.
[36] F. Scarselli, M. Gori, A. C. Tsoi, M. Hagenbuchner及G. Monfardini, 作者, 「The Graph Neural Network Model」, IEEE Transactions on Neural Networks, 卷 20, 期 1, 頁 61–80, 1月 2009, doi: 10.1109/TNN.2008.2005605.
[37] S. Wu, Y. Tang, Y. Zhu, L. Wang, X. Xie及T. Tan, 作者, 「Session-based Recommendation with Graph Neural Networks」, AAAI, 卷 33, 頁 346–353, 7月 2019, doi: 10.1609/aaai.v33i01.3301346.

[38] S. Wu, F. Sun, W. Zhang及B. Cui, 作者, 「Graph Neural Networks in Recommender Systems: A Survey」, arXiv:2011.02260 [cs], 4月 2021, 引見於: 2021年10月4日. [線上]. 載於: http://arxiv.org/abs/2011.02260
[39] C. Xu等, 作者, 「Graph Contextualized Self-Attention Network for Session-based Recommendation」, 收入 Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, Macao, China, 8月 2019, 頁 3940–3946. doi: 10.24963/ijcai.2019/547.
[40] Z. Wang, W. Wei, G. Cong, X.-L. Li, X.-L. Mao及M. Qiu, 作者, 「Global Context Enhanced Graph Neural Networks for Session-based Recommendation」, 收入 Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, Virtual Event China, 7月 2020, 頁 169–178. doi: 10.1145/3397271.3401142.
[41] F. Yu, Y. Zhu, Q. Liu, S. Wu, L. Wang及T. Tan, 作者, 「TAGNN: Target Attentive Graph Neural Networks for Session-based Recommendation」, 收入 Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, Virtual Event China, 7月 2020, 頁 1921–1924. doi: 10.1145/3397271.3401319.
[42] R. Qiu, Z. Huang, J. Li及H. Yin, 作者, 「Exploiting Cross-session Information for Session-based Recommendation with Graph Neural Networks」, ACM Trans. Inf. Syst., 卷 38, 期 3, 頁 1–23, 6月 2020, doi: 10.1145/3382764.
[43] J. Tang, M. Qu, M. Wang, M. Zhang, J. Yan及Q. Mei, 作者, 「LINE: Large-scale Information Network Embedding」, Proceedings of the 24th International Conference on World Wide Web, 頁 1067–1077, 5月 2015, doi: 10.1145/2736277.2741093.
[44] L. Lu, Y. Shin, Y. Su及G. E. Karniadakis, 作者, 「Dying ReLU and Initialization: Theory and Numerical Examples」, CiCP, 卷 28, 期 5, 頁 1671–1706, 6月 2020, doi: 10.4208/cicp.OA-2020-0165.
[45] W.-C. Kang及J. McAuley, 作者, 「Self-Attentive Sequential Recommendation」, 收入 2018 IEEE International Conference on Data Mining (ICDM), Singapore, 11月 2018, 頁 197–206. doi: 10.1109/ICDM.2018.00035.
[46] J. Tang及K. Wang, 作者, 「Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding」, 收入 Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining, Marina Del Rey CA USA, 2月 2018, 頁 565–573. doi: 10.1145/3159652.3159656.
[47] X. He, L. Liao, H. Zhang, L. Nie, X. Hu及T.-S. Chua, 作者, 「Neural Collaborative Filtering」, 收入 Proceedings of the 26th International Conference on World Wide Web, Perth Australia, 4月 2017, 頁 173–182. doi: 10.1145/3038912.3052569.

簡易檢索 / 詳目顯示

相關論文