摘要
针对新闻浏览者的偏好易变等特点,通过度量在线用户的点击和阅读行为,依据其不同的阅读策略类型,分析其页面偏好,并综合各页面偏好和新闻偏好,以关键字偏好表的形式表示;然后设计自适应的评分推荐机制,动态地分析用户兴趣及其转移;设计学习机制,根据用户实际阅读的新闻,调整其关键字偏好,并采用模糊相似度来分析用户偏好结构与新闻结构的相似性,从而产生推荐。实验表明,所构造的模型能够提供良好的个性化新闻推荐服务。
According to the characteristics of Web news browser, such as inconstancies of preference, user＇s online behaviors are measured. Firstly, user＇s page preference and news preference are analyzed at the basis of user＇s reading strategies to form a table of keywords-preferences. Then, the adaptive recommended mechanism is designed to deal with the changes of user＇s preference. Learning mechanism is also designed to adjust the keywords-preferences of user, based on news actually read by users. At last, fuzzy method is applied to analyze similarity between user＇s preference and news structure to produce recommendations. The proposed model has been approved to have higher capability than traditional method.
出处
《图书情报工作》
CSSCI
北大核心
2007年第6期77-80,71,共5页
Library and Information Service
基金
国家自然科学基金项目"面向电子商务的顾客偏好分析与个性化推荐系统"(项目编号:70402009)研究成果之一
关键词
用户行为
需求偏好
个性化推荐
学习策略
user＇s behavior requirement preference personalized recommendation learning strategy
