Abstract
With the growth of Internet of Thing (IoT) services, the discovery of IoT services becomes a very challenging issue because of its diverse and dynamic nature. IoT services' providers compete to deliver efficient and high-quality services with a variety of tasks where each service has different description. In this article, we describe the design of an IoT services discovery architecture to improve ways on services discovery in the IoT environments. In particular, we propose a service classification model that uses Center Profile Vector (CPV) based on the modified N-gram and centroid classifier to classify IoT services. The model make the most of calculating the term weight based on tfsc, dfsc in order to sort the terms in the IoT services description. In addition, we present a distance similarity method for the N-gram that helps with the difference in representation lengths between classes and IoT services descriptions. The proposed model has been evaluated using a prototype implementation and experimentations using a real-world IoT network dataset. The evaluation results offer promising classification rate in comparison with some other models. Finally, we utilized distinctive term weighting plans to build up the cross breed CPV model dependent on centroid classifiers and the altered N-gram to improve the classification execution. It is clear for the evaluation results that our new approach is better than traditional methods by around 30%.