基于混合型的Web实时推荐模型研究

摘要

针对以往个性化网站实时推荐系统存在很难预测用户未来浏览页面的不足,提出了一个混合型的实时推荐模型.该模型将动态模糊聚类技术和改进的关联规则相结合,既挖掘用户与页面的相似度权值形成知识库,又考虑用户的访问事务集增量构造访问模式树,通过修剪其相关分枝,快速生成候选推荐集,由推荐引擎附加在请求页面的底部,在不干扰用户的访问同时,又将用户感兴趣的内容推荐给用户.实验结果表明,该方法能有效地提高推荐的精确率和覆盖率以及综合评价指标.%Aimed at personalized web site real-time recommendation systems in the past are difficult to predict future user browsing the page, a mixed type real-time recommendation is proposed. The model connects the dynamic fuzzy clustering with improved rules concerned, not only combining both users and page similarity mining weight to form a knowledge base, but also considering the user's access sequence features to configure access mode tree, trimming some related branches to form candidate recommendation set quickly, which is attached to the request page bottom by the recommendation engine, without disturbing the user' s access meanwhile recommending the interesting contents to the users. Experiments show that the method can effectively improve the precision and coverage as well as comprehensive evaluation index in the recommendation.