藉由資料導向的特徵轉換改進流形正則化單類別協同過濾模型

Improving One-class Collaborative Filtering with Manifold Regularization by Data-driven Feature Transformation

10.6342/NTU201601737

連彥傑

流形正則化 ； 特徵轉換 ； 單類別協同過濾 ； 推薦系統 ； 貝氏個人化排序 ； Manifold Regularization ； Feature Transformation ； One-class Collaborative Filtering ； Recommender System ； Bayesian Personalized Ranking

國立臺灣大學資訊工程學系學位論文

2016年

碩士

鄭卜壬

英文

流形正則化可以幫助協同過濾模型引進使用者或物品的特徵作為推薦的考量。然而，大部份研究都是基於特定領域的先備知識或直覺來選擇使用的特徵，而非有正式的理論推導去連結推薦與特徵的關係。此外，特徵間的相依關係也需要被考慮。雖然許多降低維度的方式可以在將特徵轉換到較低維度的空間時，考量到特徵的重要性及相依關係，但現存方式是基於特徵的語意做轉換，而非考慮特徵與使用者回饋之間的關係。在這篇論文中，我們提出了一個資料導向的框架來訓練特徵轉換函數。基於流形正則化的概念，也就是相似的特徵會帶來相似行為的假設，我們的方法藉由基於物品的推薦方式以及優化排序的框架，從使用者的回饋資料中，為物品的特徵學習適當的特徵轉換函數。在實驗中，我們呈現出對於流形正則化的模型，藉由我們的方法轉換後的特徵，能比原始特徵或是基於語意的特徵轉換更好的提昇推薦的品質。此外，我們也展示出此方法作為推薦演算法的表現，能直接反映出特徵的品質。

Manifold Regularization is introduced to the field of recommender system for combining additional features into collaborative filtering. However, most works choose features to incorporate intuitively or by domain knowledge. Besides, the effect of feature combination needs to be considered. Although some methods of dimension reduction help to transform raw features to latent space with combination for usage, these approaches cannot give an appropriate representation for the problem of recommendation because they are based only on the semantic of features. In this work, we design a data-driven framework for training of transformation function. For the idea of manifold regularization that similar features bring similar behavior, our approach uses an item-based method and a ranking framework to optimize the representation function of item's feature from user's feedback data. In the experiment, we show that for manifold regularized model, the transformed features through the proposed method can boost performance better for recommendation problem compared with raw feature and semantic-based transformed features. Besides, we also show our framework's capability of detecting quality of feature for recommendation task.

1. [1] M. Belkin, P. Niyogi, and V. Sindhwani. Manifold regularization: A geometric framework for learning from labeled and unlabeled examples. Journal of Machine Learning Research, 7:2399–2434, 2006.
   連結：
2. [2] L. Du, X. Li, and Y.-D. Shen. User graph regularized pairwise matrix factorization for item recommendation. In Proceedings of the 7th International Conference on Advanced Data Mining and Applications - Volume Part II, ADMA’11, pages 372–385, 2011.
   連結：
3. [3] Y. Gu, B. Zhao, D. Hardtke, and Y. Sun. Learning global term weights for content-based recommender systems. In Proceedings of the 25th International Conference on World Wide Web, WWW ’16, pages 391–400. International World Wide Web Conferences Steering Committee, 2016.
   連結：
4. [4] R. He and J. McAuley. VBPR: visual bayesian personalized ranking from implicit feedback. In AAAI Conference on Artificial Intelligence, 2016.
   連結：
5. [6] I. Jolliffe. Principal component analysis. Wiley Online Library, 2002.
   連結：
6. [7] Y. Koren, R. Bell, and C. Volinsky. Matrix factorization techniques for recommender systems. Computer, 42(8):30–37, 2009.
   連結：
7. [9] R. Pan, Y. Zhou, B. Cao, N. N. Liu, R. Lukose, M. Scholz, and Q. Yang. One-class collaborative filtering. In 2008 Eighth IEEE International Conference on Data Mining, pages 502–511, 2008.
   連結：
8. [5] Y. Hu, Y. Koren, and C. Volinsky. Collaborative filtering for implicit feedback datasets. In Proceedings of the 2008 Eighth IEEE International Conference on Data Mining, ICDM ’08, pages 263–272. IEEE Computer Society, 2008.
9. [8] H. Ma, D. Zhou, C. Liu, M. R. Lyu, and I. King. Recommender systems with social regularization. In Proceedings of the Fourth ACM International Conference on Web Search and Data Mining, WSDM ’11, pages 287–296. ACM, 2011.
10. [10] S. Rendle, C. Freudenthaler, Z. Gantner, and L. Schmidt-Thieme. Bpr: Bayesian personalized ranking from implicit feedback. In Proceedings of the Twenty-Fifth Conference on Uncertainty in Artificial Intelligence, UAI ’09, pages 452–461. AUAI Press, 2009.
11. [11] Y. Shi, A. Karatzoglou, L. Baltrunas, M. Larson, N. Oliver, and A. Hanjalic. Climf: Learning to maximize reciprocal rank with collaborative less-is-more filtering. In Proceedings of the Sixth ACM Conference on Recommender Systems, RecSys ’12, pages 139–146. ACM, 2012.
12. [12] A. van den Oord, S. Dieleman, and B. Schrauwen. Deep content-based music recommendation. In Advances in Neural Information Processing Systems 26, pages 2643–2651. 2013.
13. [13] X. Wang and Y. Wang. Improving content-based and hybrid music recommendation using deep learning. In Proceedings of the 22Nd ACM International Conference on Multimedia, MM ’14, pages 627–636. ACM, 2014.
14. [14] C. Zhang, K. Wang, E. peng Lim, Q. Xu, J. Sun, and H. Yu. Are features equally representative? a feature-centric recommendation, 2015.
15. [15] J. Zheng, J. Liu, C. Shi, F. Zhuang, J. Li, and B. Wu. Dual similarity regularization for recommendation. In Advances in Knowledge Discovery and Data Mining - 20th Pacific-Asia Conference, PAKDD 2016, Auckland, New Zealand, April 19-22, 2016, Proceedings, Part II, pages 542–554, 2016.