Δημιουργία γνώσης βάσει δεδομένων μεγάλης κλίμακας για την γνωσιακή διαχείριση τηλεπικοινωνιακών υποδομών
Knowledge generation from telecommunications big data for enabling cognitive infrastructure management
View/ Open
Keywords
Συστήματα γνωσιακής διαχείρισης ; Δημιουργία γνώσης ; Μηχανική μάθηση ; Δεδομένα μεγάλης κλίμακας ; Μη καθοδηγούμενες τεχνικές μάθησης ; Cognitive Radio Systems (CRSs) ; Cognitive learning ; Knowledge building ; Machine learning ; Big data ; Unsupervised learningAbstract
The continuously growing use of Internet and the optimization of the services, in terms of offering more capabilities to the users, result in the increased need for spectrum/bandwidth, a rather limited resource, and processing capabilities in core and access networks. To this end, Cognitive Radio Systems (CRSs) have been proposed for enhancing the resource allocation and utilization, and thus bridge this gap while preserving, if not enhancing, the Quality of Services (QoS) and the Quality of Experience (QoE).
Moreover, the availability of large amounts of unstructured data, which come from various sources, is seen as highly promising for deriving high level information and new insights for the business world while easier access to them through the Web facilitates the research towards this direction. However, the velocity of them being changed requires exceptional technology to efficiently process large quantities of data within tolerable timeframes. Data characterized by high volume, variety and velocity are commonly known as Big Data. These data need to be efficiently managed, handled and exploited by the Network Operators (NOs) and/or Service Providers (SPs) but human resources are not sufficient.
Knowledge building mechanisms are often proposed for addressing both of the above challenges. In particular, cognitive network management can offer solutions to the challenges posed by future networks but this requires the incorporation of knowledge that is dynamically built from its own mechanisms. Dynamically built knowledge exploits context information and allows quicker and more complex data analysis so as to better comply with the volume, the velocity and the variety of the produced Big Data. In order to build knowledge that enhances the decisions of the network, the network monitors its current state and senses information with respect to the context it functions, it collects information regarding the results of its decisions – whether the state in which it evolved allows it to have better or worse performance – and is dynamically trained to select the state with the highest performance when in similar context. During the decision making process, rules and policies of the NO and/or the SP are combined with the knowledge built from the past experience of the network so as to be respected. To this end, this dissertation studies, designs, proposes and evaluates knowledge building mechanisms that can exploit (Big) data and enhance the decision making processes of a CRS.