Algorithm for IoT service characterization: performance evaluation

With the explosion of the number and variety of connected objects, it is necessary to define a system of recommendation of IoT services to the users so that they better use the connected objects that are at their disposal. The purpose of this recommendation system is to provide the user with customized IoT services and to help him choose suitable connected objects for the composition of these services. In this context, it is fundamental to be able to characterize IoT services to identify classes of services. We have previously proposed an algorithm based on the physical interfaces of the connected objects that allows this characterization. In this article, we evaluate its performance. To do this, we developed a data generation tool for IoT services, which was then used to evaluate the characterization algorithm. The analysis of the results allows us to conclude that our algorithm gives a good characterization of the IoT services, but sometimes additional information based on the context will be necessary in order to improve its performance.