面向最终用户的组合服务推荐  

End-user oriented composite service recommendation

作　　者：公帅[1,2] 熊锦华[1] 刘志勇[1] 

机构地区：[1]中国科学院计算技术研究所,北京100190 [2]中国科学院大学,北京100049

出　　处：《计算机集成制造系统》2013年第8期1876-1882,共7页Computer Integrated Manufacturing Systems

基　　金：国家973计划资助项目(2007CB310805);国家科技支撑计划资助项目(2011BAH11B02);国家自然科学基金创新群体资助项目(60921002)~~

摘　　要：为了提高服务组合的效率,提出基于混合相似度策略的组合服务推荐方法,基于用户未完成的组合服务来推荐相似度最高的若干已完成的组合服务。根据组合服务完成程度的不同,提出两种组合服务相似性度量方法:基于BCR模型的行为相似性度量和基于组合服务描述文档的词频—逆向文件频率内容相似性度量。实验表明,该方法的推荐准确率和推荐效率均比经典的组合服务相似性度量方法高,与人工标注数据的相关程度也较高;同时,行为相似性度量更适合完成程度更高的组合服务推荐,而内容相似性度量更适合初始阶段。To improve the efficiency oI compostte servme, a ~e, v larity strategy was proposed. Several highest similarity complete composite services were recommended based on in- complete composite services. According to the different degree of composite services, two measurement methods of composite service similarity which included a behavior similarity measurement based on Behavior Constraint Relation （BCR） model and a content similarity measurement based on Term Frequency-Inverse Document Frequency （TF- IDF） were proposed. The experiment result showed that the proposed method＇s recommendation precession and effi- ciency was higher than classical composite service, and the eorre-lation of manual annotation was also higher. The behavior similarity measurement was more suitable {or composite service with higher completeness, and the content similarity was more suitable for initial stage.

关 键 词：最终用户 组合服务 业务流程 相似度计算 服务推荐 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论]