Collaborative QoS prediction with context-sensitive matrix factorization

• Published in: Future generation computer systems 2018-05, Vol.82, p.669-678
• Main Authors: Wu, Hao, Yue, Kun, Li, Bo, Zhang, Binbin, Hsu, Ching-Hsien
• Format: Article
• Language: English
• Subjects: Cloud services; Context-sensitive; Matrix factorization; QoS prediction
• ISSN: 0167-739X; 1872-7115
• DOI: 10.1016/j.future.2017.06.020

How to obtain personalized quality of cloud/IoT services and assist users selecting the appropriate service has become a hot issue with the explosion of services on the Internet. Collaborative QoS prediction is proposed to address this issue by borrowing ideas from recommender systems. However, there is still a challenging problem as how to incorporate contextual factors into existing algorithms to realize context-aware QoS prediction as contextual factors play a crucial role in QoS assessment. In this paper, we propose a general context-sensitive matrix-factorization approach (CSMF) to make collaborative QoS prediction. By considering the complexity of service invocations, CSMF models the interactions of users-to-services and environment-to-environment simultaneously, and make full use of implicit and explicit contextual factors in the QoS data. Experimental results show that CSMF significantly outperforms the-state-of-art methods in metric of prediction accuracy. Particularly, when the QoS data is very sparse, CSMF is more effective and robust. •This paper presents a general context-sensitive approach for collaborative QoS prediction.•Interactions of users-to-services and environment-to-environment are considered simultaneously as contextual factors in the QoS data.•Our method takes advantages of both implicit and explicit factors entailed in the QoS data through exploiting contextual information.•Experimental results reflect that this study offers an efficient global optimization, enabling robust and accurate prediction results.