Prediction of Erlang-2 Distributed Primary User Traffic for Dynamic Spectrum Access

In this letter, we propose prediction of primary user spectrum activity by constructing a continuous-time Markov chain model for Erlang-2 distributed on/off channel utilization intervals. Moreover, we propose a maximum-likelihood estimator for traffic parameters utilized by the predictor. Finally, we analyze the prediction confidence as the probability of the prediction performance that is not affected by traffic estimation errors. The prediction confidence is quantified by exploiting the proposed prediction region analysis.