Performance Analysis of Cooperative Spectrum Sensing using Regression Algorithms

Abstract

Recently, cooperative spectrum sensing (CSS) became an emerging technology due to its advantages in using the spectrum efficiently. In this paper CSS is integrated with the Machine Learning (ML) algorithms to improve the sensing of users which is possible with ML algorithms only in estimating the channel status. Further, the various popular machine learning models for regression are discussed with relevant mathematical analysis such as Linear Regression, Nonlinear Regression, Gaussian Process Regression (GPR), Support Vector Machine Regression (SVM), Generalized Linear Model, Regression Tree, Shallow Neural Network, Deep Neural Network and Regression Tree Ensembles. In this paper, 20 primary users are assumed with different energy levels by maintaining a maximum energy of 20 units and a threshold value of 4 units is fixed for all four scenarios and obtained results are analyzed. In this regard, we concluded a best fit line to extract the features such that the linear regression has been chosen. In addition, various transformations also applied for further improvement in regression to obtain the best results.