基于双向关联规则项目评分预测的推荐算法研究  

Recommendation Algorithm Based on Item Rating Prediction Using Bidirectional Association Rules

作　　者：刘枚莲[1] 刘同存[1] 张峰[1] 

机构地区：[1]桂林电子科技大学商学院,桂林541004

出　　处：《武汉理工大学学报》2011年第9期150-155,共6页Journal of Wuhan University of Technology

基　　金：国家自然科学基金(70862001)

摘　　要：针对用户评分数据稀疏性问题,通过对事务数据库项目空间关联性分析,提出基于双向关联规则项目评分预测的推荐算法。算法利用双向关联规则挖掘事务数据库中相互关联的项目,找到目标项目的关联集,利用已评分项目初步预测用户对目标项目的偏好程度,最后结合协同过滤算法为用户提供推荐服务。实验结果表明,当置信度水平在60%～90%之间变动,支持度在5%～10%之间变化时,关联规则数目随着置信度和支持度水平的增加而逐渐减少,而推荐精度逐步提高。为了验证算法的有效性,选取置信度为80%以及支持度为7%与传统的推荐算法比较,所设计的算法能够较精确地找到目标项目的关联集,推荐精度和效率明显优于传统的推荐算法。A new recommendation algorithm based on item rating prediction using bidirectional association rules was designed to reduce the sparsity of user rating data through the analysis of item＇s spatial association in transaction database.Bidirectional association rules were used in this algorithm to mine the interrelated items in transaction database to find out the association set of target item,and then to predict the rate of target item by the rated items in association set.Finally,collaborative filtering algorithm was used together to provide recommendation services for users.When the confidence level changes from 60% to 90% and the support level changes from 5% to 10%,the experimental results show that the number of association rules decreased gradually with the increase of the confidence support level,and the recommendation accuracy has been improved gradually.In order to compare with traditional recommendation algorithm,the confidence level 80% and support level 7% were chosen to continue our experiment.The result shows that the algorithm designed in this paper can more accurately find out the association set of target item,the recommendation efficiency and accuracy are both obviously superior to traditional recommendation algorithm.

关 键 词：电子商务 推荐算法 双向关联规则 空间关联性 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]