Uma proposta para classificação de recursos da IoT explorando o tratamento da incerteza na decisão multicritério

Abstract

The Internet of things (IoT) is characterized by networked resources, most of which offer more than one service. The classification process for selecting the resource whose services best meet a client’s request constitutes a research front with several current and internationally relevant challenges. Among the challenges we highlight the high scalability, the dynamism with which the resources enter and leave the environment, as well as the client’s uncertainties when specifying their preferences. Thus, due to this high scalability of the IoT, a client request can return hundreds, or even thousands of resources that meet the specifications of the request considering its functional parameters. In turn, the literature review made in this Thesis identified the evaluation of non-functional parameters as a recurring technique in the various proposals for classification of resources. Considering this scenario, this Thesis has as main objective to present a proposal for classification of discovered resources in an autonomous way, called EXEHDA-RR, which is based on conception: (i) a decision approach that considers multiple criteria; and (ii) the treatment of uncertainty when defining the different parameters considered. Another aspect contemplated in this Thesis, is related to the minimization of the computational costs involved in the procedure of classification of resources in dynamic environments such as IoT, where the entry and/or exit of resources in the computing environment can occur at all times. The EXEHDA-RR evaluation was performed using usage scenarios and the resource specification was extracted from the QWS dataset, which considers real resources made available on the Internet. The results obtained show that it is possible to offer clients the most suitable resources, according to the specified preference. The use of Fuzzy Logic, among other aspects, made it possible for clients to specify their preferences more comfortably, using linguistic terms such as low, medium and high, minimizing uncertainties. The results achieved with the usage scenarios validate the proposal and point to the continuity of the research in future works.