Table 2 Comparison of spectrum prediction categories
From: Statistical spectrum occupancy prediction for dynamic spectrum access: a classification
Category | Advantages | Disadvantages |
|---|---|---|
Memoryless stochastic | Low complexity | Limited to sparse spectrum usage |
source models (Section 5) | Closed form solution for sparse | Limited to single primary |
| Â | spectrum usage scenarios | user scenarios |
| Â | Easier/convenient model to adopt | May not describe real world |
| Â | Â | channel occupancy status |
Finite order Markov models (Section 6) | Expandable to various PU/SU scenarios | Require sufficient measurements |
| Â | Â | for model training |
| Â | Applicable to heavy-tailed channel traffic | Complexity depends on the |
| Â | Â | order of the model |
| Â | Higher accuracy with manageable | Â |
| Â | number of parameters | Â |
Finite order linear regression models (Section 7) | Expandable to various PU/SU scenarios | Expandability increases the number of model parameters |
| Â | Approximation of probabilistic | Require sufficient measurements for model training |
| Â | model to linear equation model | Â |