Dynamic Spectrum Sensing in Cognitive Radio Networks using ML Model

According to CISCO report threefold increase in the mobile traffic can be witnessed between 2016 and 2021. With this, there will be a problem of spectrum allocation and usage. So, to overcome the spectrum scarcity problem, new generation spectrum management focuses on spectrum sharing and the promising technology that assists dynamic spectrum sharing is cognitive radio (CR). Spectrum sensing techniques plays an important and challenging role in spectrum sharing by avoiding the collision occurring between the primary users (PUs) and the second users (SUs). It also reduces delay and energy consumption of the SUs and base station. Machine learning algorithms are proven to be more efficient than the narrow band spectrum sensing techniques in terms of speed, complexity etc. Hence, this paper developed an algorithm for spectrum prediction process in CR adapted by gradient boosting algorithm to predict the status of the channel as busy, idle and middle. From the results, it is shown that the proposed scheme can predict the channel status effectively with a high accuracy of 99%.