Machine Learning Techniques and Models for Situation Awareness of IoT based Complex Systems

The current reality is characterized by a solid technological and pervasive component. These elements are expressed through smart devices, which make the environments we live in pervasive and able to exchange information. An example is represented by Smart Cities, complex environments able to leverage large amounts of data from sensors based on the Internet of Things (IoT) paradigm. One of the current challenges is using this information to transform scenarios from complex to helpful for increasing human well-being. This objective can be achieved by acquiring Context-Awareness, analyzing information, and managing the environment through the Situation-Awareness paradigm. This Thesis aims to introduce a methodology with predictive capabilities and context adaptability for managing complex scenarios. The added value of the proposed approach is the introduction of the semantic value acquired from the Context and Situation Awareness through graph approaches, which, unlike many strategies used, leads to better integration of knowledge, obtaining higher system performance. In particular, a methodology for merging Ontologies, Context Dimension Trees, and probabilistic approaches based on Bayesian Networks will be presented to help experts and end-users handle events and provide suggestions for improving the liveability of smart complex scenarios. The proposed methodology has been validated and applied to several complex scenarios based on the IoT paradigm obtaining promising results. [edited by Author]