已知的面向排序的协同过滤算法主要有两个缺点:计算用户相似度时只考虑用户对同一产品对的偏好是否一致,而忽略了用户对产品对的偏好程度以及该偏好在用户间的流行度; 进行偏好融合和排序时需要中间步骤来构建价值函数然后才能利用贪婪算法产生推荐列表。为解决上述问题: 我们利用类TF-IDF加权策略对用户的偏好程度及偏好流行度进行综合考量,使用加权的Kendall Tau相关系数计算用户间的相似度;进行偏好融合与排序时则使用基于投票的舒尔茨方法直接产生推荐列表。在两个电影数据集上,本文提出的算法在评测指标NDCG上的效果要明显优于其他流行的协同过滤算法。
Abstract
Most ranking-oriented collaborative filtering (CF) algorithms have two limitations. Firstly, they only consider the accordance of user preferences but ignore the degrees and popularities of user preferences when computing user similarities. Secondly, an intermediate step is necessary to formulate the value function, for preference prediction and aggregation in greedy algorithms to derive recommendation lists. To address these problems, we propose a Degree-Popularity weighting scheme integrating TF-IDF to weight the degrees and popularities of user pairwise relative preferences, and compute similarities between users based on weighted Kendall Tau rank correlation coefficient. Preference aggregations and predictions are directly formulated and the recommendation lists are consequently derived by applying the Schulze method. We conduct extensive experiments on two movie datasets under NDCG evaluation, implying advantageous results in comparison with the state-of-the-art CF algorithms.
关键词
协同过滤 /
面向排序 /
加权KendallTau /
舒尔茨方法
{{custom_keyword}} /
Key words
collaborative filtering /
ranking-oriented /
weighted Kendall Tau /
Schulze method
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Su X, Khoshgoftaar TM. A survey of collaborative filtering techniques[J]. Adv. in Artif. Intell. 2009,2009:2-2.
[2] Hu Y, Koren Y, Volinsky C. Collaborative Filtering for Implicit Feedback Datasets[C]//Proceedings of the 2008 Eighth IEEE International Conference on Data Mining: IEEE Computer Society; 2008: 263-272.
[3] Liu NN, Yang Q. EigenRank: a ranking-oriented approach to collaborative filtering[C]//Proceedings of the 31st annual international ACM SIGIR conference on research and development in information retrieval. Singapore, Singapore: ACM; 2008: 83-90.
[4] Sarwar B, Karypis G, Konstan J, Riedl J. Item-based collaborative filtering recommendation algorithms[C]//Proceedings of the 10th international conference on World Wide Web. Hong Kong, Hong Kong: ACM; 2001: 285-295.
[5] 刘建国, 周涛, 汪秉宏. 个性化推荐系统的研究进展[J]. 自然科学进展,2008,19:15.
[6] Adomavicius G, Tuzhilin A. Toward the next generation of recommender systems: a survey of the state-of-the-art and possible extensions[J]. Knowledge and Data Engineering, IEEE Transactions, 2005,17:734-749.
[7] Weimer, Markus, Karatzoglou, et al. COFIRANK-Maximum Margin Matrix Factorization for Collaborative Ranking[C]//Proceedings of NIPS; 2007.
[8] Shi Y, Larson M, Hanjalic A. List-wise learning to rank with matrix factorization for collaborative filtering[C]//Proceedings of the fourth ACM conference on recommender systems. Barcelona, Spain: ACM; 2010: 269-272.
[9] Cohen W W, Schapire R E, Singer Y. Learning to order things[J]. Artif. Int. Res. 1999,10:243-270.
[10] Wang S, Sun J, Gao B J, et al. Adapting Vector Space Model to Ranking-based Collaborative Filtering[C]//Proceedings of CIKM; 2012.
[11] Gunawardana A, Shani G. A Survey of Accuracy Evaluation Metrics of Recommendation Tasks[J]. Mach. Learn. Res. 2009,10:2935-2962.
[12] Valizadegan H J R, Zhang R, Mao J. Learning to Rank by Optimizing NDCG Measure[C]//Proceedings of the Conference on Neural Information Processing Systems (NIPS); 2009.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
基金
国家自然科学基金(61272240,60970047,61103151),教育部人文社科基金(12YJC630211),教育部博士点基金(20110131110028),山东省自然科学基金(ZR2012FM037),山东省优秀中青年科学家科研奖励基金(BS2012DX012)和山东大学研究生自主创新基金(YZC12084)
{{custom_fund}}
PDF(2173 KB)
