CAJ | 학술논문

针对传统的Markov链模型不能有效的表征长串访问序列所蕴含的丰富的用户行为特征(用户类别特征、访问兴趣迁移特征)的缺点,提出混合隐Markov链浏览模型.混合隐Markov链模型使用多个不同的模型来区分不同类别用户的浏览特征,并为每个类别的用户设置了能跟踪捕捉其访问兴趣变化的类隐Markov链模型,能更好地对WWW长串访问序列的复杂特征进行建模,在真实WWW站点访问日志数据上的用户聚类实验与个性化推荐实验的结果表明,混合隐Markov链模型与传统的Markov链模型相比,具有更理想的聚类性能和推荐性能.
침대전통적Markov련모형불능유효적표정장천방문서렬소온함적봉부적용호행위특정(용호유별특정、방문흥취천이특정)적결점,제출혼합은Markov련류람모형.혼합은Markov련모형사용다개불동적모형래구분불동유별용호적류람특정,병위매개유별적용호설치료능근종포착기방문흥취변화적류은Markov련모형,능경호지대WWW장천방문서렬적복잡특정진행건모,재진실WWW참점방문일지수거상적용호취류실험여개성화추천실험적결과표명,혼합은Markov련모형여전통적Markov련모형상비,구유경이상적취류성능화추천성능.
Since the Markov Chain Model can not denote the abundant users' behavioral characteristics(such as: characteristics of users' type, characteristics of users' interests transfer ) of a long access sequence effectively, the Mixtures of Hidden Markov Chain Models is proposed. Mixtures of Hidden Markov Chain Models use different models to distinguish the browsing categories of users from different types, and set a Hidden Markov Chain Models (can track and catch the changes of users' interests) for each users' type. Mixtures of Hidden Markov Chain Models can model the complex characteristics of the WWW long access sequences better. The results of users clustering experiment and personalized recommendation experiment with a real WWW web access log data show that Mixtures of Hidden Markov Chain Models have more perfect clustering and recommendation performance than Markov Chain Model.