基于链接预测模型的移动用户偏好预测方法的研究与实现
|
摘要
最近几年,伴随着移动终端硬件设备和软件系统等飞速发展,人们对于移动通信网的需求与日俱增。随着移动终端设备自身的信息传输、承载能力的不断提升,以及微博、社交网络和网上商城等在移动领域的逐渐盛行,移动通信网在为人们带来丰富多彩的移动网络信息的同时,也产生了日趋严重的移动通信网络信息过载等问题。用户偏好预测是缓解传统互联网中信息过载问题的重要方法之一,所以在移动通信网领域,用户偏好预测也成为缓解信息过载问题的首要途径。然而,与传统互联网相比,移动通信网有其自身的特点,如存在移动网络资源比较有限,移动终端输入输出功能的限制等,如何从浩如烟海的移动信息海洋中发掘移动用户所真正感兴趣的信息,提高移动通信网个性化服务质量,对于移动用户偏好预测的研究提出了更高的要求。
针对移动通信网所独具的特点和日益增长的移动通信网个性化服务需求,为了获取更准确的移动用户偏好,本文在用户信任度计算、链接预测和用户偏好的时间衰减特性三个方面对移动用户偏好预测相关问题进行了研究和总结。首先,通过分析移动用户的通信行为,提出了一种计算移动用户信任度的方法,并引入到用户偏好预测之中。再次,在根据确定的近似邻居集合预测用户偏好之前,通过链接预测技术确定所要进行预测的移动应用集合,在一定程度上缓解了传统的协同过滤算法的扩展性问题。然后,基于用户偏好的时间衰减特性,通过时间衰减函数模拟用户偏好的衰减趋势,并给出了改进后的用户偏好预测方法。最后实验证明,改进后的方法在缓解了传统协同过滤技术的数据稀疏性问题和扩展性问题的同时,也在一定程度上提高了移动用户偏好预测的精度。
In recent years, with the quick development of mobile terminal both in hardware and software, there is an increasing requirement for mobile communication networks. Mobile terminal is becoming more popular for its convenience as well as the increasing transmission and load-bearing capability. What's more, twitter, social networks and online shopping is prevailing during this period, and our life is becoming awesome for various of mobile communication network resources. Meanwhile, it brings us a deteriorating mobile information overloading problem as well. User preference prediction is one of the most important ways to alleviate information overloading problem, so it becomes the primary way when people handle this problem in the field of mobile communication networks. Compared with traditional Internet, there are less network recourses in mobile communication networks and the I/O ability of the devices which are used to receiving mobile information is also limited. So it is becoming crucial to find out what the mobile users really want in the vast recourse ocean so as to enhance the quality of personalized mobile service. And all above raise a higher requirement for the research work of mobile user preference prediction.
In order to obtain more accurate mobile user preferences with the characteristics of mobile communication networks and the increasing personalized service requirement. In this paper we study and summarize mobile user preference through trust degree, link prediction and time decay. Firstly, we propose a mobile user trust calculation method through analyzing user communication behavior and put it into mobile user preference prediction. And we ascertain the mobile applications which are to be predicted before user preference prediction according to the nearest neighbors of the active users. Then we imitate the decay tendency with some decay functions according the decay feature of user preference and propose an improved user preference prediction method. Experimental results show that the improved method can obtain more accurate user preferences compared with traditional collaborative filtering as well as mitigating data sparsity problem and extensibility problem.
针对移动通信网所独具的特点和日益增长的移动通信网个性化服务需求,为了获取更准确的移动用户偏好,本文在用户信任度计算、链接预测和用户偏好的时间衰减特性三个方面对移动用户偏好预测相关问题进行了研究和总结。首先,通过分析移动用户的通信行为,提出了一种计算移动用户信任度的方法,并引入到用户偏好预测之中。再次,在根据确定的近似邻居集合预测用户偏好之前,通过链接预测技术确定所要进行预测的移动应用集合,在一定程度上缓解了传统的协同过滤算法的扩展性问题。然后,基于用户偏好的时间衰减特性,通过时间衰减函数模拟用户偏好的衰减趋势,并给出了改进后的用户偏好预测方法。最后实验证明,改进后的方法在缓解了传统协同过滤技术的数据稀疏性问题和扩展性问题的同时,也在一定程度上提高了移动用户偏好预测的精度。
In recent years, with the quick development of mobile terminal both in hardware and software, there is an increasing requirement for mobile communication networks. Mobile terminal is becoming more popular for its convenience as well as the increasing transmission and load-bearing capability. What's more, twitter, social networks and online shopping is prevailing during this period, and our life is becoming awesome for various of mobile communication network resources. Meanwhile, it brings us a deteriorating mobile information overloading problem as well. User preference prediction is one of the most important ways to alleviate information overloading problem, so it becomes the primary way when people handle this problem in the field of mobile communication networks. Compared with traditional Internet, there are less network recourses in mobile communication networks and the I/O ability of the devices which are used to receiving mobile information is also limited. So it is becoming crucial to find out what the mobile users really want in the vast recourse ocean so as to enhance the quality of personalized mobile service. And all above raise a higher requirement for the research work of mobile user preference prediction.
In order to obtain more accurate mobile user preferences with the characteristics of mobile communication networks and the increasing personalized service requirement. In this paper we study and summarize mobile user preference through trust degree, link prediction and time decay. Firstly, we propose a mobile user trust calculation method through analyzing user communication behavior and put it into mobile user preference prediction. And we ascertain the mobile applications which are to be predicted before user preference prediction according to the nearest neighbors of the active users. Then we imitate the decay tendency with some decay functions according the decay feature of user preference and propose an improved user preference prediction method. Experimental results show that the improved method can obtain more accurate user preferences compared with traditional collaborative filtering as well as mitigating data sparsity problem and extensibility problem.
引文
[1]曹建平.探析移动通信技术的发展趋协[J].无线互联科技,2010,(1)：22-23.
[2]吴泽世.移动通信网技术标准与业务应用[J].重庆通信业,2010,(6)：35-37.
[3]康乐,荆继武,王跃武等.社会化网络服务中的信任扩张和控制[J].计算机研究与发展,2010,47(9)：1611-1621.
[4]余学军.六度分割理论成就SNS[J]信息网络,2008,(11)：37.
[5]周宇煜.移动社会化网络业务发展趋势和商业模型探讨[C].中国通信学会无线及移动通信委员会、IP应用与增强电信技术委员会2007年度联合学术年会论文集.北京,2007：65-96.
[6]Hclgc L,Thomas N.A latent model for collaborative filtering[J].International journal of approximate reasoning,2012,53(4):447-466.
[7]李国,张智斌,刘芳先等.非线性组合的协同过滤推荐算法[J].计算机应用,2011,31(11)：3063-3067.
[8l张光卫,李德毅.李鹏等.基于云模型的协同过滤推荐算法[J].2007,18(10)：2403-2411.
[9]曾小波.魏祖宽,金在弘等.协同过滤系统的矩阵稀疏性问题的研究[J].计算机应用,2010,30(4)：1079-1082.
[10]罗喜军,王韬丞,杜小勇等.基于类别的推荐——一种解决协同推荐中冷启动问题的方法[C].第二十四届中国数据学术会议论文集.2007：290-295.
[11]王小斌,矫立峰,孙延明等.基于模糊模拟的主观信任评价模型研究[J].计算机工程与用,2008,44(16)：102-104.
[12]乔秀全,杨春,李晓峰等.社交网络服务中一种基于用户上下文的信任度计算方法[J].计算机学报,2011,34(12)：2403-2413.
[13]明庆华,王明雷.教育信任：学校交际文化建设的基石[J].现代教育论丛,2010,(11)：72-76,63.
[14]王勇,代桂平,侯亚荣等.基于贝叶斯网络的组合服务信任度评估方法[J].高技术通讯,2010,20(1)：21-25.
[15]李想,王宇,张建伟等.一种改进的公钥证书抗攻击信任度模型[J].计算机应用,2007,27(8)：1919-1921,1925.
[16]鲍美英,高玉斌,马礼等.网络环境下基于域的信任模型[J].计算机测量与控制,2010.18(1)：102-194.197.
[17]司冠南,任宇涵,许静等.基于贝叶斯网络的网构软件可信性评估模型[J].计算机研究 与发展,2012,49(5)：1028-1038.
[18]李小勇,桂小林.大规模分布式环境下动态信任模型研究[J].软件学报,2007,18(6)：1510-1521.
[19]黄武汉,孟祥武,王立才.移动通信网中基于用户社会化关系挖掘的协同过滤算法[J].电子与信息学报,2011,33(12)：3002-3007.
[20]暴世宏,江春先.边际效用递减规律的再发现[J].价值工程,2012,31(17)：120-123.
[21]Huang D. and Arasan V. On measuring email-based social network trust[C]. In:Proc. of the Global Telecommunications Conference (GLOBECOM 2010), Miami, USA,2010.
[22]Avesani P.,Massa P. and Tiella R.. A Trust-Enhanced Recommender System Application: Moleskiing[C]. In:Proceeding SAC '05 Proceedings of the 2005 ACM symposium on Applied computing. New York, USA,2005.
[23]刘玉龙,曹元大,李剑等.一种新型推荐信任模型[J].计算机工程与应用,2004,40(29)：47-49,80.
[24]王玉祥,乔秀全,李晓峰等.上下文感知的移动社交网络服务选择机制研究[J].计算机学报,2010,33(11)：2126-2135
[25]Adomavicius G. and Tuzhilin A.. Towards the next generation of recommender systems:a survey of the state-of-the-art and possible extensions[J]. IEEE Transactions on Knowledge and Data Engineering,2005,17(6):734-749.
[26]Eagle N., Pentland A., and Lazer D.. Inferring friendship network structure by using mobile phone data[C]. Proceedings of the National Academy of Sciences,2009,106(36):15274.
[27]Benchettara N., Kanawati R. and Rouveirol C.. Supervised machine learning applied to link prediction in bipartite social networks [C]. International Conference on Advances in Social Networks Analysis and Mining,2010:326-330.
[28]郭景峰,王春燕,邹晓红等.一种改进的针对合著关系网络的链接预测方法[J].计算机科学,2008,35(12)：126-128.
[29]郭景峰,代军丽,马鑫等.针对通信社会网络的时间序列链接预测算法[J].计算机科学与探索,2010,04(6)：552-559.
[30]Huang Zan, Li Xin, and Chen Hsinchun. Link prediction approach to collaborative filtering[C]. In:Proc. of the 5th ACM/IEEE-CS Joint Conference on Digital Libraries, New York, USA,2005.
[31]袁尧,张玉成,董支霞等.基于二分图匹配的多业务流网络选择机制[J].软件学报,2()10,21(6)：1378-1390.
[32]饶君,吴斌,东昱晓MapReduce环境下的并行复杂网络链路预测.软件学报,2012,23(12)：3175-3186
[33]Michael Mitzenmacher. A brief history of lognormal and power law distributions[C]. In: Proceedings of the Allerton Conference on Communication. Control, and Computing. 2001: 182-191.
[34]Leo Katz. A new status index derived from sociometric analysis[J]. Psychomctrika, 1953, 18(1):39-43.
[35]李燕,陈莹,董秀兰等.基于神经网络的遥感图像识别算法[J].测绘与空间地理信息,2012,35(2)：156-158.
[36]唐亚伟,秦玉平.基于数据挖掘的分类算法综述[J].渤海大学学报,2011,32(4)：372-375.
[37]奉国和.SVM与神经网络在时间序列预测中的比较[J].现代管理科学,2006,(9)：40-41.
[38]B Li. SJ Song, K Li et al. Improved conjugate gradient implementation for least squares support vector machines[J]. Pattern recognition letters. 2012, 33(2): 121-125.
[39]Su JS. Zhang BF, Xu X. Advances in machine learning based text categorization[J]. Journal of Software, 2006, 17(9): 1X48-1X59.
[40]李方,赵英凯,颜昕等.基于Matlab的最小二乘支持向量机的工具籍及其应用[J].计算机应用,2006,26(2)：358-360.
[41]王健,唐新怀.移动中间件中上下文相关的对象模型[J].计算机应用与软件,2007,24(5)：63-64,74
[42]印莹,刘志勇,武晓新等.基于上下文的Web服务推荐模型[J].广西师范大学学报,2009,27(1)：133-136.
[43]Wolfgang Wocrndl, Johann Schlichtcr. Introducing Context into Rccommcnclcr Syslcms[C]. The Fifth Workshop on Intelligent Techniques for Web Personalization and the Workshop on Rccommcndcr Systems in K-Commcrcc. Garching, Germany, 2007: 138-140.
[44]徐风苓,孟祥武,王立才.基于移动用户上下文相似度的协同过滤推荐算法[J].电子与信息学报,2011,33(11)：2785-2789
[45]Wang I.C, Menu XW and Zhang YJ. A heuristic approach to social network-based and conicxt-awnrc mobile services recommendation |J|. Journal of Convergence Information Technology, 2011, 6( 10): 339-346.
[46]于洪,李转运.基于遗忘曲线的协同过滤推荐算法[J].南京大学学报,2010,46(5)520-527.
[47]Lei Li. Li Zheng and Tao Li. A long-short user interest integration in personalized news rccommcndation|C|. In: Proc. of the 5th ACM/IEEE-CS Joint Conference on Digital Libraries. New York, USA, 2005.
[48]G. Gim, H. Jcong, H. Lee and D. Yun. Group-aware prediction with exponential smoothing for collaborative filtering[C] In: Proceeding CAMR' II Proceedings of the 2nd Challenge on Context-Aware Movie Recommendation, New York, USA,2011:11-14.
[49]Shi YC, Meng XW, Wang LC. A heuristic approach to identifying the specific household member for a given rating[C]. In:Proc. of the 2nd Challenge on Context-Aware Movie Recommendation. New York, USA,2011:47-52.
[501 Li Xue and Ding Yi. Time weight collaborative filtering[C]. In:Proceeding CIKM '05 Proceedings of the 14th ACM international conference on Information and knowledge management, New York, USA,2005:485-492.
[51]李佳伦,谷利泽,杨义先等.一种具有时间衰减和主观预期的P2P网络信任管理模型[J].电子与信息学报,2009,31(11)：2786-2790.
[2]吴泽世.移动通信网技术标准与业务应用[J].重庆通信业,2010,(6)：35-37.
[3]康乐,荆继武,王跃武等.社会化网络服务中的信任扩张和控制[J].计算机研究与发展,2010,47(9)：1611-1621.
[4]余学军.六度分割理论成就SNS[J]信息网络,2008,(11)：37.
[5]周宇煜.移动社会化网络业务发展趋势和商业模型探讨[C].中国通信学会无线及移动通信委员会、IP应用与增强电信技术委员会2007年度联合学术年会论文集.北京,2007：65-96.
[6]Hclgc L,Thomas N.A latent model for collaborative filtering[J].International journal of approximate reasoning,2012,53(4):447-466.
[7]李国,张智斌,刘芳先等.非线性组合的协同过滤推荐算法[J].计算机应用,2011,31(11)：3063-3067.
[8l张光卫,李德毅.李鹏等.基于云模型的协同过滤推荐算法[J].2007,18(10)：2403-2411.
[9]曾小波.魏祖宽,金在弘等.协同过滤系统的矩阵稀疏性问题的研究[J].计算机应用,2010,30(4)：1079-1082.
[10]罗喜军,王韬丞,杜小勇等.基于类别的推荐——一种解决协同推荐中冷启动问题的方法[C].第二十四届中国数据学术会议论文集.2007：290-295.
[11]王小斌,矫立峰,孙延明等.基于模糊模拟的主观信任评价模型研究[J].计算机工程与用,2008,44(16)：102-104.
[12]乔秀全,杨春,李晓峰等.社交网络服务中一种基于用户上下文的信任度计算方法[J].计算机学报,2011,34(12)：2403-2413.
[13]明庆华,王明雷.教育信任：学校交际文化建设的基石[J].现代教育论丛,2010,(11)：72-76,63.
[14]王勇,代桂平,侯亚荣等.基于贝叶斯网络的组合服务信任度评估方法[J].高技术通讯,2010,20(1)：21-25.
[15]李想,王宇,张建伟等.一种改进的公钥证书抗攻击信任度模型[J].计算机应用,2007,27(8)：1919-1921,1925.
[16]鲍美英,高玉斌,马礼等.网络环境下基于域的信任模型[J].计算机测量与控制,2010.18(1)：102-194.197.
[17]司冠南,任宇涵,许静等.基于贝叶斯网络的网构软件可信性评估模型[J].计算机研究 与发展,2012,49(5)：1028-1038.
[18]李小勇,桂小林.大规模分布式环境下动态信任模型研究[J].软件学报,2007,18(6)：1510-1521.
[19]黄武汉,孟祥武,王立才.移动通信网中基于用户社会化关系挖掘的协同过滤算法[J].电子与信息学报,2011,33(12)：3002-3007.
[20]暴世宏,江春先.边际效用递减规律的再发现[J].价值工程,2012,31(17)：120-123.
[21]Huang D. and Arasan V. On measuring email-based social network trust[C]. In:Proc. of the Global Telecommunications Conference (GLOBECOM 2010), Miami, USA,2010.
[22]Avesani P.,Massa P. and Tiella R.. A Trust-Enhanced Recommender System Application: Moleskiing[C]. In:Proceeding SAC '05 Proceedings of the 2005 ACM symposium on Applied computing. New York, USA,2005.
[23]刘玉龙,曹元大,李剑等.一种新型推荐信任模型[J].计算机工程与应用,2004,40(29)：47-49,80.
[24]王玉祥,乔秀全,李晓峰等.上下文感知的移动社交网络服务选择机制研究[J].计算机学报,2010,33(11)：2126-2135
[25]Adomavicius G. and Tuzhilin A.. Towards the next generation of recommender systems:a survey of the state-of-the-art and possible extensions[J]. IEEE Transactions on Knowledge and Data Engineering,2005,17(6):734-749.
[26]Eagle N., Pentland A., and Lazer D.. Inferring friendship network structure by using mobile phone data[C]. Proceedings of the National Academy of Sciences,2009,106(36):15274.
[27]Benchettara N., Kanawati R. and Rouveirol C.. Supervised machine learning applied to link prediction in bipartite social networks [C]. International Conference on Advances in Social Networks Analysis and Mining,2010:326-330.
[28]郭景峰,王春燕,邹晓红等.一种改进的针对合著关系网络的链接预测方法[J].计算机科学,2008,35(12)：126-128.
[29]郭景峰,代军丽,马鑫等.针对通信社会网络的时间序列链接预测算法[J].计算机科学与探索,2010,04(6)：552-559.
[30]Huang Zan, Li Xin, and Chen Hsinchun. Link prediction approach to collaborative filtering[C]. In:Proc. of the 5th ACM/IEEE-CS Joint Conference on Digital Libraries, New York, USA,2005.
[31]袁尧,张玉成,董支霞等.基于二分图匹配的多业务流网络选择机制[J].软件学报,2()10,21(6)：1378-1390.
[32]饶君,吴斌,东昱晓MapReduce环境下的并行复杂网络链路预测.软件学报,2012,23(12)：3175-3186
[33]Michael Mitzenmacher. A brief history of lognormal and power law distributions[C]. In: Proceedings of the Allerton Conference on Communication. Control, and Computing. 2001: 182-191.
[34]Leo Katz. A new status index derived from sociometric analysis[J]. Psychomctrika, 1953, 18(1):39-43.
[35]李燕,陈莹,董秀兰等.基于神经网络的遥感图像识别算法[J].测绘与空间地理信息,2012,35(2)：156-158.
[36]唐亚伟,秦玉平.基于数据挖掘的分类算法综述[J].渤海大学学报,2011,32(4)：372-375.
[37]奉国和.SVM与神经网络在时间序列预测中的比较[J].现代管理科学,2006,(9)：40-41.
[38]B Li. SJ Song, K Li et al. Improved conjugate gradient implementation for least squares support vector machines[J]. Pattern recognition letters. 2012, 33(2): 121-125.
[39]Su JS. Zhang BF, Xu X. Advances in machine learning based text categorization[J]. Journal of Software, 2006, 17(9): 1X48-1X59.
[40]李方,赵英凯,颜昕等.基于Matlab的最小二乘支持向量机的工具籍及其应用[J].计算机应用,2006,26(2)：358-360.
[41]王健,唐新怀.移动中间件中上下文相关的对象模型[J].计算机应用与软件,2007,24(5)：63-64,74
[42]印莹,刘志勇,武晓新等.基于上下文的Web服务推荐模型[J].广西师范大学学报,2009,27(1)：133-136.
[43]Wolfgang Wocrndl, Johann Schlichtcr. Introducing Context into Rccommcnclcr Syslcms[C]. The Fifth Workshop on Intelligent Techniques for Web Personalization and the Workshop on Rccommcndcr Systems in K-Commcrcc. Garching, Germany, 2007: 138-140.
[44]徐风苓,孟祥武,王立才.基于移动用户上下文相似度的协同过滤推荐算法[J].电子与信息学报,2011,33(11)：2785-2789
[45]Wang I.C, Menu XW and Zhang YJ. A heuristic approach to social network-based and conicxt-awnrc mobile services recommendation |J|. Journal of Convergence Information Technology, 2011, 6( 10): 339-346.
[46]于洪,李转运.基于遗忘曲线的协同过滤推荐算法[J].南京大学学报,2010,46(5)520-527.
[47]Lei Li. Li Zheng and Tao Li. A long-short user interest integration in personalized news rccommcndation|C|. In: Proc. of the 5th ACM/IEEE-CS Joint Conference on Digital Libraries. New York, USA, 2005.
[48]G. Gim, H. Jcong, H. Lee and D. Yun. Group-aware prediction with exponential smoothing for collaborative filtering[C] In: Proceeding CAMR' II Proceedings of the 2nd Challenge on Context-Aware Movie Recommendation, New York, USA,2011:11-14.
[49]Shi YC, Meng XW, Wang LC. A heuristic approach to identifying the specific household member for a given rating[C]. In:Proc. of the 2nd Challenge on Context-Aware Movie Recommendation. New York, USA,2011:47-52.
[501 Li Xue and Ding Yi. Time weight collaborative filtering[C]. In:Proceeding CIKM '05 Proceedings of the 14th ACM international conference on Information and knowledge management, New York, USA,2005:485-492.
[51]李佳伦,谷利泽,杨义先等.一种具有时间衰减和主观预期的P2P网络信任管理模型[J].电子与信息学报,2009,31(11)：2786-2790.