融合停留时间的隐Markov个性化推荐模型

doi:10.3969/j.issn.1000-436x.2014.09.011

通信学报 ›› 2014, Vol. 35 ›› Issue (9): 112-121.doi: 10.3969/j.issn.1000-436x.2014.09.011

融合停留时间的隐Markov个性化推荐模型

刘胜宗¹,樊晓平^1,²,廖志芳³,胡佳¹

¹ 中南大学信息科学与工程学院，湖南长沙 410075
² 湖南财政经济学院网络化系统研究所，湖南长沙 410205
³ 中南大学软件学院，湖南长沙 410075

出版日期:2014-09-25 发布日期:2017-06-14
基金资助:
国家科技支撑计划基金资助项目;国家自然科学基金资助项目;湖南省自然科学基金资助项目;湖南省科技支撑计划基金资助项目

Hidden Markov model fused with staying time for personalized recommendation

Sheng-zong LIU¹,Xiao-ping FAN^1,²,Zhi-fang LIAO³,Jia HU¹

¹ School of Information Science and Engineering,Central South University,Changsha 410075,China
² Laboratory of Networked Systems,Hunan University of Finance and Economics,Changsha 410205,China
³ School of Software,Central South University,Changsha 410075,China

Online:2014-09-25 Published:2017-06-14
Supported by:
The National Key Technology R＆D Program of China;The National Natural Science Foun-dation of China;The Natural Science Foundation of Hunan Province;Key Technology R＆D Program of Hu-nan Province

摘要/Abstract

摘要：

静态模型在推荐系统中往往将用户的兴趣偏好看作是固定不变的，而在一定程度上与实际并不符合。为此，基于隐Markov动态模型提出一种融合停留时间的类时齐隐Markov个性化推荐模型（ctqHMM）。该模型用隐含状态变量的转移来模拟 Web 用户的兴趣变迁，并用停留时间来描述用户对某一偏好感兴趣的程度和所推荐页面的重要性。然后，提出一种基于该模型平稳分布的用户聚类方法，并将其用于推荐系统中。在真实的 Web服务器访问记录数据上的实验证明，类时齐隐Markov模型具有更好的推荐性能。

关键词: Web挖掘, 类时齐隐Markov模型, 平稳分布, 用户聚类, 个性化推荐, HMM

Abstract:

Static model in the recommendation system often regards the user's interest as changeless,which is inconsis-tent with the actual to a certain extent.With regards to this,a hidden Markov model fused with staying time for personal-ized recommendation (ctqHMM) based on the HMM dynamic model is proposed.The proposed model employs the transfer of the implicit state variables to simulate the changes of Web users' interests,and uses staying time to describe the level of interest to the specific preference and the importance of the recommended pages.Then,a user's clustering method based on the stationary distribution of the ctqHMM is also proposed and applied into the recommending systems.Experiment results on real Web server access log data show the encouraging performance of the proposed method over the state-of-the-arts.

Key words: Web mining, classified time homogeneous hidden Markov model, stationary distribution, user clustering, personalized recommendation, HMM

刘胜宗,樊晓平,廖志芳,胡佳. 融合停留时间的隐Markov个性化推荐模型[J]. 通信学报, 2014, 35(9): 112-121.

Sheng-zong LIU,Xiao-ping FAN,Zhi-fang LIAO,Jia HU. Hidden Markov model fused with staying time for personalized recommendation[J]. Journal on Communications, 2014, 35(9): 112-121.

图/表 8

图1

图2

表1

图3

图4

表2

表3

图5

参考文献 19

[1]	AWAD M , ISSA K . Prediction of user's web-browsing behavior:application of markov model[J]. IEEE Transactions on Systems,Man,and Cybernetics,Part B: Cybernetics, 2012,42(4):1131-1142.
[2]	SAHOO N , VIR S , MUKHOPADHYAY T . A hidden Markov model for collaborative filtering[J]. MIS Quarterly-Management Information Systems, 2012,36(4):13-29.
[3]	许海玲，吴潇等．互联网推荐系统比较研究[J]．软件学报， 2009,20(2):350-362. XU H L , WU X , et al. Comparison study of Internet recommendation system[J]. Journal of Software, 2009,20(2):350-362.
[4]	BHAWNA N , SURESH J . Generating a new model for predicting the next accessed web page in web usage mining[A]. Proceeding in 3rd International Conference on Emerging Trends in Engineering and Technology[C]. India, 2010.485-490.
[5]	MIKA P . Ontologies are us: a unified model of social networks and semantics[J]. Web Semantics: Science,Services and Agents on the World Wide Web, 2007,5(1):5-15.
[6]	OARD DW , KIM J . Implicit feedback for recommender systems[A]. Proceeding of the AAAI Workshop on Recommender Systems[C]. USA, 1998.81-83.
[7]	BORGES J , LEVENE M . Data mining of user navigation patterns[A]. Proceedings of the 1999 KDD Workshop on Web Mining[C]. San Diego California, 1999.92-111.
[8]	邢永康，马少平．多Markov 链用户浏览预测模型[J]．计算机学报， 2003,26(11):1510-1517. XING Y K , MA S P . Modeling user navigation sequences based on multi-markov chains[J]. Chinese Journal of Computers, 2003,26(11):1510-1517.
[9]	LIU Y T , LIU T Y , GAO B , et al. Browserank: letting web users vote for page importance[A]. Proceedings of the 31st Annual International ACM SIGIR Conference on Research and Development in Informa-tion Retrieval[C]. Singapore, 2008.451-458.
[10]	JOSEPH A , JOHN R . Recommender systems: from algorithms to user experience[J]. User Model User-Adap Inter, 2012,22:101-123.
[11]	LICHTENWALTER R N , LUSSIER J T , CHAWLA N V . New per-spectives and methods in link prediction[A]. Proceedings of the 16th ACM SIGKDD International Conference on Knowledge Discovery and Data mining[C]. USA, 2010.243-252.
[12]	KNIJNENBURG B P , WILLEMSEN M C , GANTNER Z , et al. Explaining the user experience of recommender systems[J]. User Model User-Adap Inter, 2012,22:441-504.
[13]	刘凯鹏，方滨兴．一种基于社会性标注的网页排序算法[J]．计算机学报， 2010,33(6):1014-1023. LIU K P , FANG B X . A novel page ranking algorethm based on social annotations[J]. Chinese Journal of Computers, 2010,33(6):1014-1023.
[14]	HOTHO A , JASCHKE R , SCHMITZ C , et al. Information Retrieval in Folksonomies: Search and Ranking[M]. Berlin Heidelberg, Springer, 2006.411-426.
[15]	SONG G H , SUN S T , FAN W . Applying user interest on item-based recommender system[A]. Proceedings of the 5th International Joint Conference on Computational Sciences and Optimization (CSO)[C]. Harbin,China, 2012.635-638.
[16]	边肇祺，张学工等．模式识别(第二版)[M]．北京：清华大学出版社， 2000. BIAN Z Q , ZHANG X G , et al. Pattern Recognition(2)[M]. Beijing: Tsinghua University Press, 2000.
[17]	WAN M , JONSSON A , WANG C , et al. New Frontiers in Applied Data Mining Lecture Notes in Computer Science[J]. New Frontiers in Applied Data Mining Lecture Notes in Computer Science, 2012,7104:40-52.
[18]	孙荣恒．随机过程及其应用[M]．北京：清华大学出版社， 2004. SUN R H . Stochastic Process and Application[M]. Beijing: Tsinghua university press, 2004.
[19]	朱郁筱，吕琳媛．推荐系统评价指标综述[J]．电子科技大学学报， 2012,2(41):163-175. ZHU Y X , LV L Y . Evaluation metrics for recommender systems[J]. Journal of University of Electronic Science and Technology of China, 2012,2(41):163-175.

数据	用户数目	页面数目	请求数目
数据1	32 710	5 771	127 536
数据2	300 930	246 971	1 379 746

用例编号	L平均值(数据1)	数据1的准确率		L平均值(数据2)	数据2的准确率
用例编号	L平均值(数据1)	本文聚类方法	传统Markov聚类	L平均值(数据2)	本文聚类方法	传统Markov聚类
1	3.56	0.226 1	0.224 2	2.92	0.216 2	0.217 1
2	6.37	0.531 2	0.429 3	5.66	0.520 3	0.417 4
3	8.41	0.649 4	0.531 6	7.48	0.632 7	0.525 8
4	9.22	0.768 5	0.654 7	8.69	0.757 6	0.651 9
5	11.68	0.824 3	0.696 5	10.42	0.811 9	0.702 2
6	13.78	0.838 7	0.756 4	13.67	0.833 4	0.739 2
7	15.63	0.847 1	0.808 2	15.92	0.849 5	0.811 2
8	16.91	0.869 1	0.816 1	17.83	0.871 2	0.824 1

目标用户浏览序列长度L_test		数据1的准确率			数据2的准确率
目标用户浏览序列长度L_test	100用户(L_train=11.32)	300用户(L_train=11.94)	500用户(L_train=11.85)	100用户(L_train=13.64)	300用户(L_train=13.63)	500用户(L_train=13.79)
1	0.225 1	0.212 5	0.234 1	0.209 6	0.224 1	0.231 4
3	0.325 6	0.332 1	0.339 8	0.335 5	0.374 5	0.342 9
5	0.402 1	0.416 8	0.417 5	0.411 3	0.420 3	0.429 8
7	0.532 6	0.551 2	0.556 8	0.565 9	0.570 1	0.586 3
9	0.607 7	0.623 5	0.634 7	0.627 5	0.632 1	0.639 8
11	0.610 8	0.632 1	0.645 5	0.638 7	0.643 3	0.653 9
13	0.620 3	0.642 2	0.658 4	0.647 2	0.657 4	0.662 5

融合停留时间的隐Markov个性化推荐模型

Hidden Markov model fused with staying time for personalized recommendation

在线阅读

PDF下载

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 8

参考文献 19

相关文章 5

Metrics

推荐阅读 0

[1]	任永功,张云鹏,张志鹏. 基于粗糙集规则提取的协同过滤推荐算法[J]. 通信学报, 2020, 41(1): 76-83.
[2]	仇一泓，尧婷娟，秦丰林，葛连升. 基于用户过滤的校园无线网用户聚类方法[J]. 通信学报, 2014, 35(Z1): 28-149.
[3]	仇一泓,尧婷娟,秦丰林,葛连升. 基于用户过滤的校园无线网用户聚类方法[J]. 通信学报, 2014, 35(Z1): 146-149.
[4]	罗奇,余英,赵呈领,曹艳. 自适应推荐算法在电子超市个性化服务系统中的应用研究[J]. 通信学报, 2006, 27(11): 183-186.
[5]	郭伟刚. 电子商务数据挖掘中的预处理技术[J]. 通信学报, 2005, 26(1A): 275-278.