Personalised QoS-based web service recommendation with service neighbourhood-enhanced matrix factorisation

Abstract

Web service recommendation has been an urgent task with the wide adoption of services in cloud computing and enterprise information systems, and Quality of Service (QoS) prediction takes a critical role. In this paper, based on probabilistic matrix factorisation model, we propose three service–neighbourhood enhanced prediction models. In the first model, after identifying the neighbours sharing similar QoS records, we learn the predicted value both utilising feature vectors of the active service and its neighbours. Then, in the second model, the learning process is enhanced through the integration with each neighbour's weight computed from the similarity. Afterwards, in the final model, we emphasise the special roles of services possessed by the same company with the active service, with minimising the difference of feature vectors between them. Further, we propose a unified recommendation framework, involving in the three models. Finally, we conduct extensive experiments showing our models' effectiveness.