Figure 3. Scheme of two-level 10-fold cross-validation.

The data set is partitioned into 10 parts (folds) in the outer loop. One fold of the data set is kept for testing of SVM. The remaining 9 folds are used as the training set for training an SVM. In the inner loop, the training set is further divided into 10 folds to choose the optimal parameters for testing the accuracy of the data set kept in the outer loop. The procedure is repeated 10 times.