摘要
在基于QoS的Web服务的选择中,用户需要对未使用过的服务的QoS进行预测。目前主要利用已有的用户服务QoS信息矩阵,预测缺失的QoS。但随着网络上Web服务数量的激增,用户-服务QoS信息矩阵的稀疏性问题成为影响预测效果的重要因素。针对传统方法的不足,提出一种基于近邻数判定的预测方法,通过比较待预测项的用户邻居数和服务邻居数,自动选择采用基于用户或者基于服务的协同过滤预测方法,实验表明,该方法能有效提高QoS预测结果的精度。
The QoS of Web services should be predicted before users select the services. The history QoS data of serv- ices are usually utilized for the prediction. With the increase of service numbers, the data sparsity influences the prediction quality badly. To slove the problem, this paper propose a neighbor numbers decision based method. Through comparing the numbers of user＇s neighbors and item＇s neighbors, our approach decides the users based or the item based collaborative filtering will be applied. Experimental result shows that the quality of prediction results is largely improved.
出处
《舰船电子工程》
2014年第9期38-41,共4页
Ship Electronic Engineering
基金
总装预研基金(编号:9140A27040413JB11407)资助
关键词
WEB服务
QoS预测
邻居数
协同过滤
Web service, QoS prediction, neighbor numbers, collaborative filtering
