چكيده به لاتين
Cell outage management, including outage detection and compensation, is one of the most prominent parts of Self-organizing Networks. An outage can be caused by different external elements such as local weather, which should be detected concerning various aspects of a network. Meanwhile, outage compensation entails models that are capable of interacting with the environment actively. Consequently, in recent years, researchers have focused on Cell Outage Management (COM) using machine learning approaches. Deep reinforcement learning models significantly outperformed other alternatives on COM problems. In this thesis, we will present a framework for both outage detection and compensation. In our detection scheme, we used CVFDT as a stream classifier to detect the occurrence of outages, with 91.7 AUC outperformed K-NNAD. Furthermore, we developed a Deep Q network to compensate outage in cellular networks based on optimizing cells' transmission rate alongside co-type cells' connectivity. Our COC method has been converged for different random initialization and increasingly improved reward values, which led to support 95.81% and 96.7% of target connectivity in macro and femtocells, respectively.
