Alleviating Data Sparsity in Web Service QoS Prediction by Capturing Region Context Influence

With the advent of service computing paradigm, Web service QoS prediction has become a necessity to support high quality service recommendation and reliable Web-based system building. However, the inherent data sparsity issue and potentially strong but inconspicuous relation between users or Web services and their neighborhoods under the context of region information are overlooked in previous studies. In this paper, we propose a unified matrix factorization model by capturing the influences of region contexts from both user and service sides in an integrated way. Different from previous researches, our approach capitalizes on the advantages of latent feature and neighborhood approaches systematically so as to achieve accurate QoS prediction. Experimental results have shown the proposed approach outperforms its competitive methods with respect to accuracy efficiently, thereby demonstrating the positive effect that incorporation of explicit region context can have on alleviating the concerned data sparsity issue.