藉由深度學習的方法推薦旅程

Tour Recommendations by Deep Learning Approaches

許安然(An-Jan Hsu)

指導教授 ： 李瑞庭

國立臺灣大學/管理學院/資訊管理學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201901958

摘要

隨著分享照片的社群平台興起，越來越多使用者在這些平台上分享即時資訊和照片，個人化旅程推薦可幫助使用者規劃他們的行程，也可幫助旅行社客製化行程以及進行目標式廣告。探索有效的特徵與拜訪順序是旅程推薦的重要步驟，而深度學習的模型可從資料中萃取多層次的內在特徵，這些特徵可有效地代表使用者間的關係以及使用者與地點間的關係。因此，在本研究中，我們提出了一個深度學習的方法，整合使用者和地點的高階特徵、社群影響力以及拜訪順序的注意力機制，以推薦個人化旅遊行程。實驗結果顯示，我們的方法無論在推薦精準度或是平均分數方面，都勝過目前最先進的方法。我們所提出的方法，可有效地利用照片分享社群平台上的資料進行旅遊行程推薦，也可強化社群推薦系統的設計，亦可提升個人化旅遊行程推薦的品質。

關鍵字

旅遊推薦 ； 照片分享社群網站 ； 深度學習 ； 循環神經網絡 ； 資料探勘

並列摘要

With the increasing popularity of photo sharing social networks, more and more users share their timely information and photos on these platforms with their friends. Personalized tour recommendation is important for helping users plan their trips and for helping tour agents customize travelling tours for their customers and launch targeted advertising. Discovering effective features and visiting orders from check-in data is the key to tour recommendation. Deep learning models can extract multiple levels of intrinsic features from input data, which can better represent the relationships between users, and between user and point-of-interest (POI). Therefore, in this study, we propose a deep learning approach to integrate high-level features of users and POIs, social influence, and an attention mechanism of visiting order for tour recommendation. The experimental results show that our proposed approach outperforms other state-of-the-art methods in terms of hit rate and average score. Theoretically, our study contributes to the effective usage of data and analytics for social recommender system design. In practice, our results can be used to improve the quality of personalized tour recommendation services.

並列關鍵字

Tour recommendation ； photo sharing social network ； deep learning ； recurrent neural network ； data mining

參考文獻

[1] A. Ajorlou, A. Jadbabaie and A. Kakhbod, Dynamic pricing in social networks: The word-of-mouth effect, Management Science, Vol. 64, No. 2, pages 971–979, 2018.

[2] R. Baral, T. Li and X. Zhu, CAPS: Context aware personalized POI sequence recommender system, arXiv:1803.01245, pages 1-17, 2018.

[3] R. M. Bell, Y. Koren and C. Volinsky, Modeling relationships at multiple scales to improve accuracy of large recommender systems, Proceedings of the 13th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 95-104, 2007.

[4] D. M. Blei, A. Y. Ng and M. I. Jordan, Latent Dirichlet allocation, Journal of Machine Learning Research, Vol. 3, pages 993-1022, 2003.

[5] I. Brilhante, J. A. Macedo, F. M. Nardini, R. Perego and C. Renso, Where shall we go today? Planning touristic tours with tripbuilder, Proceedings of the 22nd ACM International Conference on Information and Knowledge Management, pages 757-762, 2013.