一种用于服务质量预测的混合协同过滤方法  

A Hybrid Collaborative Filtering Approach for Service Qo S Prediction

作　　者：苏凯[1,2] 马良荔[2] 刘霞[3] 郭晓明[2] 

机构地区：[1]海军工程大学装备经济管理系,湖北武汉430033 [2]海军工程大学计算机工程系,湖北武汉430033 [3]海军工程大学训练部,湖北武汉430033

出　　处：《武汉理工大学学报（信息与管理工程版）》2015年第2期207-211,共5页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：总装预研基金资助项目(9140A27040413JB11407)

摘　　要：不同用户的网络状况、异构环境和地理位置的不同导致其对相同Web服务感受到的Qo S差异较大,因此提出了一种为用户预测其未调用过服务的Qo S的混合协同过滤方法。首先采用皮尔逊相关系数计算服务之间的相似度,并引入相似度信任因子对相似度进行修正,然后得到与目标服务最相似的top-k服务近邻集,最后采用改进的协同过滤方法预测目标服务的Qo S值。采用网络上真实服务的Qo S数据集进行实验评估,实验结果表明,该方法可有效提升Qo S预测的准确度。The QoS of web service may vary widely for different users due to their different network conditions, heterogeneous environments and locations.In this paper, a hybrid collaborative filtering approach was proposed for QoS prediction.Firstly, Pearson Correlation Coefficient was employed to measure the similarity between two services to find the top-k similar neighbors for target service.Then, the similarity was modified by employing similarity trustworthy factor.Finally, the traditional collabora-tive filtering method was modified to predict the QoS of target service.Real world QoS data were obtained to evaluate the ap-proach.The experimental results demonstrate that the prediction accuracy of QoS could be effectively improved.

关 键 词：WEB服务 服务选择 服务质量预测 协同过滤 皮尔逊相关系数 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]