*3.5. The Results of SLR and SVM*

This section analyzes the effects of two other classical machine learning methods (i.e., SLR and SVM), as described in Sections 2.4 and 2.5. Figure 10 shows scatter plots of the true and estimated wind speeds. Table 8 shows the retrieval performance of each regression trees model. Similarly, the bold font represents the best result, and the performance of high wind speed models was analyzed in three data intervals. It is obvious that the retrieval results of SVM had less dispersion than those of SLR, which means that the performance of SVM was better. However, the retrieval performance of the models described in the previous two sections was better than that of the models presented in this section.

**Figure 9.** (**a**) ANN scatter plot with the activation function Sigmoid in the low wind speed interval. (**b**) ANN scatter plot with the activation function Tanh in the high wind speed interval.

**Figure 10.** Results of wind speed retrievals based on SLR and SVM.


**Table 8.** Retrieval performance of SLR and SVM.
