基于Web日志的实时推荐系统

Real-time Recommendation System Based on Web Log

针对个性化实时推荐系统的不足,提出通过构造BP树的方法压缩访问事务集。采用一个实时推荐的系统模型,将耗时的数据预处理放在离线模块,实时推荐采用动态修剪BP树的方法,穿过访问模式树的相关部分,利用网页推荐算法得到频繁访问集,生成推荐集。结果表明该算法只需扫描数据库一次,得到的频繁模式可满足页面实时推荐的快速需求。

Aiming at the insufficiency of personalized real-time recommendation system, this paper proposes the theory through constructing the BP tree method, and compressing visit business collection, using a Real-Time Recommendation System（RTRS） model, putting time-consuming data pre-processing module on the off-line one, recommending the use of dynamic BP tree pruning method in real-time, passing through the visit of the relevant parts of the pattern tree, obtaining the frequent visit collection by way of the homepage recommendation algorithm so as to produce recommendation collection. Result indicates that this algorithm only needs to scan the database one time, the frequent pattern obtained can meet the rapid demands of the Web page recommendation in real-time.