An Algorithm to Generate Synthetic Dataset for Modern Intelligent Spectrum Sensing in Cognitive Radio Networks

In the recent age of the digital world and the Fourth Industrial Revolution (4IR), the machine or deep learning approaches are progressively expanding in the context of data analysis and processing, which usually empowers applications to operate intelligently. An accurate, balanced, adaptive, and reliable machine or deep learning model fully depends on the particularly effective training dataset. This paper proposes an algorithm to prepare a synthetic dataset for applying the machine or deep learning approaches in conventional spectrum sensing techniques for cognitive radio networks. In this work, the K nearest neighbor (KNN) supervised machine learning algorithm is implemented on the generated dataset to check the robustness of the proposed dataset preparation algorithm. The proposed method reaches a classification accuracy of 98.95% at an SNR level of - 10 dB.