How Does Neighborhood Matter? An Empirical Study on MF Based QoS Prediction Framework (Short Paper)

摘要

Quality-of-Service (QoS) is a fundamental element in Service-Oriented Computing (SOC) domain. Sufficient and accurate QoS information benefits to the parts not just in-depth analysis in theory but also in system deployments in practice. Studies of predicting missing QoS values is gaining important as an essential step to fast support operations in both academia and industry. Although there are a number of ongoing research works concerning how to build up Matrix Factorization (MF) based QoS prediction frameworks, it is usually not so easy for researchers and architects to adopt these tools, since there is a paucity of research literature on the general construction methodology and significant factors affecting performance. In this paper, we provide lessons by empirically examining how the neighborhood effect, the key concept of Collaborative Filtering (CF), matters to a robust MF based QoS engine. This work first proposes a general construction methodology to integrate neighborhood factor, and then evaluates significant elements matter to the final accuracy. In doing so, practitioners can clearly follow distilled guidelines for the successful implementation of neighborhood integrated QoS prediction frameworks, and make extensions for further purposes.