CoMFS: A Collaborative Matrix Factorization System for Quality-of-Service Prediction (Short Paper)

The study on sufficient QoS data contributes to advancing the state in Service-Oriented Computing (SOC) industry and academy. To collect a large amount of resource in practice, QoS prediction systems are designed. Nevertheless, how to build up an accurate and efficient system is still a challenging problem. This paper proposes CoMFS, a Collaborative Matrix Factorization System that leverages users´ geographical information and history records to tackle the problem of Quality-of-Service (QoS) Prediction, with emphasis on achieving higher performance in relatively lower latency. To simultaneously achieve these two significant goals, CoMFS first proposes a two-stage neighborhood selection mechanism that can identify a set of highly relevant local neighbors for each target user. And then, CoMFS collects the wisdom of crowds to build up an extended Matrix Factorization (MF) engine for personalized QoS prediction. Finally, CoMFS iteratively generates results by utilizing hints from previous round computations, a memo boosting strategy that directly accelerates solving process in response to practical requests. Experimental evidence on large-scale real-world QoS data shows that CoMFS is scalable, efficient, and consistently outperforming other state-of-the-art approaches in prediction accuracy.