基于关联规则的标签推荐  

Tag Recommendation Based on Association Rules

作　　者：陈双双 王晓军[1] CHEN Shuang-shuang;WANG Xiao-jun(School of Computer Science,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学计算机学院,江苏南京210003

出　　处：《计算机技术与发展》2018年第12期43-47,共5页Computer Technology and Development

基　　金：国家自然科学基金(61003237)

摘　　要：随着数据的海量增长,为了快速筛选有用数据,许多应用系统引入了标签分类。标签在实际中得到广泛的应用,用户可以用标签标注或者搜索自己感兴趣的资源。但是,系统为用户推荐标签时常常遇到标签数据稀疏问题,无法准确捕捉用户的偏好和精准的推荐标签。为了解决这个问题,提出了一种基于关联规则的标签推荐方法。首先,为了收集用户使用过的标签数据,构建了一个重叠的滑动时间窗口模型,在这种模型上进行数据收集可以缓解数据稀疏问题;然后,分析这些标签数据,得到频繁共现的标签集集合;最后,在上述得到的频繁共现标签集集合上挖掘出标签与标签之间的关联规则。实验结果表明,利用标签之间的关联规则进行标签推荐能够有效地提高推荐的准确度。With the massive growth of data,many application systems have introduced into label classification in order to quickly screen useful data.Tag is widely used in practice.Users can label or search resources with tags.However,when the system recommends tags to users,it often exists data sparseness,which cannot accurately capture users’preferences and precise recommendation tags.To solve this problem,we present a tag recommendation method based on association rules.First,we construct an overlapping time window model to collect the tag data used.Then,analysis of the tag data we get the frequent co-existing tag set.Finally,we mine association rules between tags and labels in the base of the frequent co-existing tag set.Experiment shows that the recommendation precision is improved by using tag recommendation method based on association rules.

关 键 词：标签推荐 关联规则 稀疏性 准确度 时间窗口 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]