一种考虑用户兴趣转移特征的协同预测模型.pdfVIP

2016 年 5 月 四川大学学报(自然科学版) May. 2016 第 53 卷第3 期 Journal of Sichuan University (Natural Science Edition) Vo l. 53 No.3 doi: 103969/j. issn. 0490-6756. 2016. 05. 013 一种考虑用户兴趣转移特征的协同预测模型 刘汉清，朱 敏，苏亚博，唐彬彬 (四川大学计算机学院，成都 610065) 摘 要:大多数预测模型使用用户属性或社交关系信息来优化预测结果，然而真实系统中用 户的属性或社交关系信息往往很难获得，或者取得的是虚假信息，从而导致用户行为表达不准 确或模型不具有普适性.另外，几乎所有使用用户特征的模型仅考虑用户兴趣本身的度量，而 忽视兴趣的变化这一重要特征.因此，本文提出一种考虑用户兴趣转移特征的协同预测模型. 该模型根据用户连续行为序列构建用户兴趣转移特征和用户行为演变网络，利用用户兴趣转 移特征计算用户相似性，进而搜索最近邻集合，利用用户行为演变网络筛选候选集，最后设计 最频繁项提取算法来产生预测结果，从而构建用户行为的预测模型.在真实的新闻浏览日志、 交互式网络电视视频访问日志和微软服务器日志上的实验表明该预测模型是有效的. 关键词z 预测模型;兴趣转移;演变网络;协同过滤 中图分类号: TP182 文献标识码:A 文章编号: 0490-6756(2016)03-0548-07 A collaborative prediction model for user interest shift feature LIU Han-Qi吨， ZHU Min , 5U};α-Bo , TANG Bin-Bin (College of Computer Science , Sichuan University , Chengdu 610065 , China) Abstract: Most of prediction models employ basic attributes of users or social relationships as the origi- nal features to optimize prediction results. Nevertheless , in the real systems , its highly difficult to ac- quire attributes and social information of users , which always lack of sufficient accuracy , so that user behavior couldnt be expressed accurately as well as prediction models couldnt be applied widely. Addi- tionally , almost all the models employing user features focus on measurement of interests of user itself , while ignoring the feature of the change of user interests. Based on the above considerations , the authors propose a collaborative prediction model based on transfer features of user interests. The model con- structs transfer features of use