一种基于扩展FP-TREE的服务推荐方法  

A service recommended method based on extended FP-TREE

作　　者：莫同[1] 褚伟杰[1] 李伟平[1] 吴中海[1] 

机构地区：[1]北京大学软件与微电子学院,北京100871

出　　处：《华中科技大学学报（自然科学版）》2013年第S2期81-87,共7页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61033005);国家科技支撑计划资助项目(2012BAH06B01);高等学校博士学科点专项科研基金资助项目(20120001120119;20120001110086);深圳市科技研发资助项目(CXY201107010258A)

摘　　要：针对协作过滤推荐的矩阵稀疏性与扩展膨胀问题,提出一种基于扩展FP-TREE的改进方法.将用户的情境取值抽象为情境空间状态,通过挖掘情境状态与服务的关联进行服务推荐.引入倒排索引扩展FPTREE频繁项头表,建立状态-状态与服务-服务关联;通过索引树表示状态-服务关联,给出扩展FP-TREE与协作过滤矩阵的映射机制,在继承协作过滤的基础上极大地压缩了过滤矩阵.仿真实验表明:与传统的协作过滤推荐算法相比,该推荐方法具有更高的效率.In order to solve the sparsely and expansion problems of collaborative filtering,an improved method based on the extension of FP-TREE was proposed.States in context space was used to present context values,and service was recommended by mining association between states and service.Inverted index was involved into frequent item list to present states-states and service-service relations, and states-service was shown in index tree.Mapping mechanism from collaborative filtering matrix to extended FP-TREE was also discussed.The method could inherit compression filter and compress the matrix extremely.Simulation result shows that,compared with the traditional collaborative filtering method,the method has a higher efficiency.

关 键 词：服务推荐 情境感知 关联挖掘 协作过滤 FP-TREE 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]