摘要
针对电子商务推荐系统本质上要解决的三个问题———数据源、数据模型和推荐策略,结合最新报道的相关推荐系统,提出并在实验室条件下实现了一个推荐系统原型。为提高该推荐系统的的通用性,采用顾客购买历史这种数据源格式,而不是常见的用户评分数据;另外,为保证产生足够的推荐结果并提高其质量,用关联规则和序列规则结合的方法来构建推荐系统引擎,并设计了一个基于一次表扫描时间的推荐策略。最后,从定性和定量两方面说明该推荐系统效率高,有更好的推荐质量。
Based on the emerging relevant recommendation systems, a rule-based recommendation system prototype was implemented. Its contributions focused on coping with three essential issues to develop a recommendation system: data source, data model and recommendation strategy as a whole. To improve its flexibilities, the system employed customers' purchase histories as data source, but not the commonly-used user ratings. In addition, to ensure the quantity and quality of recommendations, its underlining engine was built up on an association-sequential-rule model. A recommendation strategy based on one-round table scanning was also designed to improve the response delay. Laboratory experiment results show that this recommendation system produces a better outcome in both efficiency and quality.
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2004年第8期898-902,共5页
Computer Integrated Manufacturing Systems
基金
广东省自然科学基金资助项目(031539)。~~
关键词
数据挖掘
规则挖掘
电子商务
推荐系统
data mining
rule mining
e-commerce
recommendation system
