摘要
将上下文推荐系统同贝叶斯网络相结合,提出了一个上下文推荐算法,并设计了上下文资源推荐系统架构。首先利用贝叶斯网络,通过计算用户访问时间和资源信息的联合概率分布来取得用户在该环境下对资源的兴趣,然后比较当前用户所处环境所选取的资源与过去环境用户选取的资源的相似度,从而为用户提供合适的资源列表。最后将所提算法同其他常用的推荐系统算法进行了比较,系统架构按照M/G/1队列进行建模,对系统架构性能和稳定性进行了验证,取得较好结果。
Combining the context-aware recommender system with Bayesian networks, we proposed a context-aware recommendation algorithm, and designed a context-aware Resource Recommendation System Architecture. First using the Bayesian network, through computing the joint probability distribution of user access time and resource information, user interest of resources in the environment was obtained, and then the similarity of current user environment selected resources and past user environment selected resources was compared to provide appropriate resources list. Finally we compared the proposed algorithm with other common recommender system algorithms, and at the same time, the system architecture was modeled according to the M/G/1 queue. The result proves the superiority and stability of our system architecture and the algorithm are better.
出处
《计算机科学》
CSCD
北大核心
2014年第7期275-278,共4页
Computer Science
基金
河南省教育厅科学技术研究重点项目:高校数字化教学资源整合与应用研究(12A520024)
河南省教育厅科学技术研究重点项目:基于社会网络的个性化微博推荐技术研究(14A520085)
河南省教育厅科学技术研究重点项目:基于网格的新型科技信息服务平台的关键技术研究(12B520016)资助
关键词
推荐系统
上下文
贝叶斯网络
期望极大化
Recommender system, Context, Bayesian network, Expectation-maximization