A statistical analysis of the measurements, not reported
here, revealed that the distribution of the prediction error is
Gaussian with good approximation. The distribution param-
eters of the prediction error for the specific street types,
when erbf function is used, are calculated and shown in
Table 2. According to the obtained results, it can be con-
cluded that SVM model A achieves a greater accuracy of the
prediction. Although the error is somehow higher in some
specific situations, total rms error is just about 3 dB greater
than the corresponding value of the local mean uncertainty.
Having in mind very high field strength dynamics (over
80 dB), it is evident that the accuracy of the proposed model
is good.