International Journal of Engineering and Technology

This paper presents the development and evaluation of three nonlinear models to forecast the power reception of different channels of the global system for mobile communications (GSM) in order to analyze spatial opportunity to reuse frequencies by secondary users (SUs) in a cognitive radio (CR) network. Markov, empirical mode decomposition-support vector regression (EMD-SVR) and wavelet neural models were utilized to forecast the channel occupancy status. Results were evaluated using the criteria of availability and occupancy times of channels, different types of mean error, and observation time. This study forecasts not only the reception power but also the occupancy and availability time of channels to determine the accuracy percentage that can have of the channel use time for the primary users (PUs) and SUs in CR systems. The analysis of the models presents the wavelet neural model as the one with the best behavior in the variables evaluated. Thus, this study suggests that this model represents a promising alternative to CR systems.