Evaluación del desempeño de modelos de decisión espectral predictivos.

This research study aims to assess the performance of two predictive spectral decision models (Naïve Bayes and logistic regression) by analyzing interference and prediction quality levels. Real spectral occupancy data are used in two types of scenarios (low traffic and high traffic), based on five metrics: 1) number of total handoffs, 2) number of failed handoffs, 3) number of handoffs with interference, 4) number of anticipated handoffs and 5) number of perfect handoffs. The results show that the Naive Bayes model performs better than the logistic regression model. However, in spite of low interference levels, it is not possible to make decisions with a single evaluation metric since high handoff rates significantly affect communication quality. It is concluded that interference and the number of channels are the two most relevant metrics. [ABSTRACT FROM AUTHOR]