With the rapid development of technologies based on Web service, a large quantity of Web services are available on the Internet. Web service recommendation aims at helping users in designing and developing service-oriented software systems. How to recommend web services with better QoS value receives a lot of attention. Previous works are usually based on the assumption that the QoS information is available. However, we usually encounter data sparsity issue, which demands the prediction of QoS Value. Also, the QoS Value of Web services may change over time due to the dynamic environment. How to handle the dynamic data streams of incoming service QoS value is a big challenge. To address above problems, we propose an Web Service Recommendation Framework by considering the temporal information. We explore to envision such QoS value data as a tensor and transform it into tensor factorization problem. A Tucker decomposition (TD) method is proposed to cope with the model which includes multidimensional information: user, service and time. To deal with the dynamic data streams of service QoS value, We introduce an incremental tensor factorization (ITF) method which is scalable, and space efficient. Comprehensive experiments are conducted on real-world Web service dataset and experimental results show that our approach exceed other approaches in efficiency and accuracy.
展开▼
