Cloud manufacturing service QoS prediction based on neighbourhood enhanced matrix factorization
Cloud manufacturing service QoS prediction based on neighbourhood enhanced matrix factorization
复制标题
DOI：
10.1007/s10845-018-1409-8
复制
发表时间：
2018-03
影响因子：
8.3
通讯作者：
Yu Feng;Biqing Huang
中科院分区：
文献类型：
--
作者：
Yu Feng;Biqing Huang
随着云制造（CMfg）的快速发展，服务质量（QoS）预测在CMfg服务平台中变得越来越重要，因为获取所有服务QoS值是不切实际的。在本文中，我们提出了一种邻域增强矩阵分解方法来预测丢失的 QoS 值。我们首先系统地考虑地理信息、样本集多样性计算和平台上下文，以扩展基本的皮尔逊相关系数（PCC）相似性并提取邻域信息。然后，我们将邻域信息整合到矩阵分解（MF）中并对缺失值进行预测。与现有方法相比，该方法具有以下新特点：（1）在计算PCC相似度时，采用熵信息来推导不同用户或服务的个人权重； (2)考虑位置信息和样本集相似度来增强PCC相似度； (3)引入拓扑信息来解决数据稀疏问题； (4)将邻域信息与MF结合以提高预测精度。我们在真实数据集上进行了实验，其中包括 339 个服务用户对 5825 个服务的 Web 服务调用，以验证我们方法的可行性和效率。
With the rapid development of cloud manufacturing (CMfg), quality-of-service (QoS) prediction becomes increasingly important in CMfg service platform because it turns out to be impractical to acquire all service QoS values. In this paper, we present a neighbourhood enhanced matrix factorization approach to predict missing QoS values. We first systematically consider geographical information, sample set diversity computation and platform context to extend basic Pearson Correlation Coefficient (PCC) similarity and extract neighbourhood information. Then, we integrate neighbourhood information into matrix factorization (MF) and make prediction of missing values. Compared with existing methods, the proposed method has the following new features: (1) entropy information is adopted to derive personal weights for different users or services when computing PCC similarity; (2) location information and sample set similarity are considered to enhance PCC similarity; (3) topology information is introduced to address data sparsity issue; (4) neighbourhood information is incorporated with MF to improve prediction accuracy. We conduct an experiment on a real-world dataset which includes web service invocations from 339 service users on 5825 services to verify the feasibility and efficiency of our method.