Quantitative Evaluation of QoS Prediction in IoT

Internet of Things (IoT) applications, are typically built from services provided by heterogeneous devices, which are potentially resource constrained and/or mobile. These services and applications are widespread and a key research question is how to predict user side quality of service (QoS), to ensure optimal selection and composition of services. Invoking all available IoT services for evaluation is impractical due to the exponential growth in these services. To assess the current state of the art related to this research question we conduct a quantitative evaluation of QoS prediction approaches, particularly those that use matrix factorisation (MF) for collaborative QoS prediction. These approaches derive from the collaborative filtering model used in recommender systems, which avoids the problems of many other QoS prediction approaches of requiring additional invoking of available IoT components on behalf of the user. We conduct comprehensive experiments based on a real-world large-scale QoS dataset as well as a transformation of this dataset to more closely estimate IoT services, to show the prediction accuracy of these approaches. We also give a demonstration of how they can be used in a small scale example.