在音樂串流服務成為行動裝置必備的應用程式時，意味著每天有無數聆聽音樂的選擇，故音樂推薦系統是留住使用者的重要服務。對供應商而言，音樂線上串流服務則可協助搜集使用者和歌曲的相關資訊，幫助其效能提升，因此音樂推薦於使用者或供應商之間的依存關係，使得音樂推薦逐漸盛行起來。
音樂推薦可考量的影響因素相當多元，包含樂曲風格、樂曲要素和種類等等，而使用者傾聽偏好的社會因素及地理位置，甚至是聆聽時的情境(scenario)以及重覆聆聽的行為也都是重要的考量之一。其難題在於，音樂數量相比電影數量的擴增速度更快，要準確地進行推薦的難度也隨之提升，另外聆聽音樂的成本低廉，但影響的因素複雜，故導致影響歌曲聆聽的動機變得碎片化，其聆聽行為相比電影觀賞行為則較為不固定、規律性較低且不一定會忠實反映使用者對歌曲的興趣所在。
本研究的資料使用KKBOX的使用者及歌曲聆聽相關資訊，預測使用者於接下來一個月內是否重複聽取某首歌曲，我們設計使用者特徵、歌曲特徵和使用者聆聽歌曲時的情境特徵，透過深度學習完成協同式過濾推薦系統，並首先對以模型為主(model-based)的矩陣分解模型以及其延伸模型(NCF、NeuMF)進行效能比較，接著讓特徵結合矩陣分解延伸模型，協助其效能提升，最後對WSDM Cup 2018音樂推薦的最佳模型(Lystdo[22])提出之特徵以及其模型進行架構及因子分析。

When music streaming service nearly becomes an essential application on mobile devices, people have thousands of choice about listening to music. Namely, music recommendation is an important service to retain customers. For content providers, they can gather click data by online streaming service to improve system performance. Therefore, music recommendation systems gradually prevail on account of the dependency relation for it between customers and content providers.
The influential factors of recommendation consist of social factors, geographic location and listening scenario from user preference as well as listening repeatedly behavior. One key problem is that the numbers of songs grow more rapidly than those of movies, which results in increasing difficulties to recommend songs accurately. Since the cost for listening music is low, the factors can be weak and fragmented which leads to complicated motivation. Comparing listening behavior with watching movies behavior, we find the former is unfixed and has low regularity. That is to say, listening behavior could not reflect perfectly the true interest to songs from users.
In this paper, we focus on the KKBOX to predict whether or not the recurring listening event will be triggered with a month. We design context features, user features and song features for combination with deep collaborative filtering. On our experiment, we compare matrix factorization models with the extended model (NCF, NeuMF). We combine features with matrix factorization extended model to improve the performance. Finally, we analyze the features factor and architecture from the best model (Lystdo[22]) in WSDM Cup 2018 music recommendation task.

[1] Jonathan Herlocker, Joseph Konstan, Loren Terveen and John Riedl. Evaluating collaborative filtering recommender systems. In SIGIR ’99, pages 230–237, 1999.
[2] Xiaoyuan Su and Taghi M. Khoshgoftaar. A survey of collaborative filtering techniques. Advances in Artificial Intelligence, pages 1–19, 2009(Section 3).
[3] Zan Huang, Hsinchun Chen, and Daniel Dajun Zeng. Applying associative retrieval techniques to alleviate the sparsity problem in collaborative filtering. ACM Trans. Information Systems, vol. 22, no. 1, pages 116–142, 2004.
[4] Michael Pazzani and Daniel Billsus. Content-based recommendation systems. The Adaptive Web, pages 325–341, 2007.
[5] Pasquale Lops, Marco de Gemmis and Giovanni Semeraro. Recommender Systems Handbook, 2011.
[6] Dmitry Bogdanov, Joan Serra, Nicolas Wack, Perfecto Herrera, and Xavier Serra. Unifying low-level and high-level music similarity measures. IEEE Transactions on Multimedia, 13(99):1–1, 2011.
[7] Pedro Cano, Markus Koppenberger, and Nicolas Wack. An Industrial-strength content-based music recommendation system. In Proceedings of the 28th annual international ACM SIGIR conference on Research and development in information retrieval - SIGIR ’05, page 673, New York, New York, USA, 2005. ACM Press.
[8] Yehuda Koren, Robert Bell, and Chris Volinsky. Matrix factorization techniques for recommender systems. IEEE Computer, 42(8):30–37, 2009.
[9] Robin Burke. Hybrid recommender systems: survey and experiments. User Modelling and User-Adapted Interaction. vol. 12, no. 4, pages 331–370, 2002.
[10] Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu and Tat-Seng Chua. Neural collaborative filtering. In WWW, pages 173–182. ACM, 2017.
[11] Yading Song, Simon Dixon, and Marcus Pearce. A survey of music recommendation systems and future perspectives. In Proceedings of the 9th International Symposium on Computer Music Modelling and Retrieval, 2012.
[12] Aaron van den Oord, Sander Dieleman, and Benjamin Schrauwen. Deep content-based music recommendation. In NIPS 26, pages 2643–2651, 2013.
[13] Hao Wang, Naiyan Wang, and Dit-Yan Yeung. Collaborative deep learning for recommender systems. In Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’15, pages 1235–1244, 2015.
[14] Sander Dieleman. Recommending music on Spotify with deep learning. http://benanne.github.io/2014/08/05/spotify-cnns.html, 2014.
[15] Paul Covington, Jay Adams, and Emre Sargin. Deep neural networks for youtube recommendations. In Proceedings of the 10th ACM Conference on Recommender Systems. ACM, 191–198, 2016
[16] Shuai Zhang, Lina Yao and Aixin Sun, Deep learning based recommender system: A survey and new perspectives. CoRR, vol. abs/1707.07435, 2017.
[17] Hanh T. H. Nguyen, Martin Wistuba, Josif Grabocka, Lucas Rego Drumond, and Lars Schmidt-Thieme. Personalized deep learning for tag recommendation. Springer International Publishing, Cham, 186–197. https://doi.org/10.1007/978-3-319-57454-7 15, 2017.
[18] Yuyun Gong and Qi Zhang. Hashtag recommendation using attention-based convolutional neural network. In IJCAI. 2782–2788, 2016.
[19] Balazs Hidasi, Alexandros Karatzoglou, Linas Baltrunas, and Domonkos Tikk. Session-based Recommendations with Recurrent Neural Networks. CoRR, abs/1511.06939, 2015, 2015.
[20] Suvash Sedhain, Aditya Krishna Menon, Scott Sanner, and Lexing Xie. Autorec: Autoencoders meet collaborative filtering. In Proceedings of the 24th International Conference on World Wide Web. ACM, 111–112, 2015.
[21] How Jing and Alexander J Smola. Neural survival recommender. In Proceedings of the Tenth ACM International Conference on Web Search and Data Mining. ACM, 515–524, 2017.
[22] Lystdo. WSDM – KKBox’s Music Recommendation Challenge. https://www.kaggle.com/c/kkbox-music-recommendation-challenge/discussion/45942