Abstract—Cognitive radio is a technology that allows a more efficient spectrum use. Currently, there are several proposals for the spectral mobility of secondary users. However, very few of them are proactive. The advantage of proactive models is that they allow a reduction in the level of interference between primary and secondary users due to the prediction of the arrival of the primary. This work presents the comparative evaluation of two predictive spectral decision models, Naïve Bayes, and Logistic Regression, analyzing the level of interference they generate and the level of quality in the prediction. The evaluation is done through five metrics in low and high traffic scenarios: failed handoffs, interference handoffs, bandwidth, delay and throughput; in two types of scenarios: low traffic and high traffic. The main contributions of this work are the use of real spectral occupancy data captured in previous measurement campaigns and the robust evaluation carried out through five metrics, among which the interference measured from the moment in which the primary and secondary users coexist in the same channel, and the analysis under a radio environment with many and few primary users. The results show that the Naïve Bayes model has a 20% better performance than the logistic regression model; this together with its low processing level makes it an excellent candidate for the selection of selecting opportunities in cognitive radio networks.
 

Keywords—benchmarking, cognitive radio networks, Multiple Criteria Decision Making (MCDM), spectrum decision, spectrum handoff, spectrum mobility