基于张量分解的动态Web服务推荐

在服务计算领域中，为了能够在大量具有相同功能的Web服务以及API等数据资源中选择适合用户的服务和接口，提出了服务推荐系统。当前常用的基于服务质量（QoS）的服务推荐系统所采用的模型假定服务的QoS值恒定不变，是一种由服务和用户的二元关系构成的二维静态模型。针对实际应用中，QoS是受到多种因素影响的变量这一问题，提出了一种可以描述多个影响QoS因素的张量模型，并利用张量分解算法来对服务推荐算法进行了改进。实验结果表明：提出的基于张量分解的服务推荐算法与6种现有算法相比较，预测服务的QoS值的绝对平均误差（MAE）不同程度地降低了20%~50%，并且所建模型能够描述更多的影响因素，从而可对服务进行动态推荐。

Dynamic Web service recommendation based on tensor factorization

In the area of Web service computing, in order to select a suitable service for users in a large number of Web services and API with the identical function，the issue of Web service recommendation is becoming more and more critical. At present, in the quality of service (QoS) based service recommendation systems, the hypothesis of the system model is a two-dimensional static model which is composed of dyadic relationship between users and service interaction. However, in view of the practical application, the QoS value is affected by many factors, and a tensor model is proposed to describe the factors which affect the QoS. Then, we propose a method to discover the latent factors that govern the associations among these multi-type objects of QoS. A new recommendation approach based on tensor factorization is proposed to address the issue of Web service QoS value prediction with considering Web service invocation time. The experimental results show that compared with six related algorithms, the mean absolute error (MAE) of the proposed tensor factorization algorithm is reduced by 20%-50%, and our model can be used to describe more factors and to dynamically recommend Web service.