Table 2.

Prediction of plant clades using nested SVMs.

Clade	Samples	Training set	Accuracy			Test set	Accuracy	Precision	Recall
			SVM1	SVM2	SVM3
Ferns	1176	919	0.84	–	–	257	0.92	0.87	0.72
Eudicots	3534	2841	0.78	0.75	–	693	0.78	0.76	0.89
Angio-/gymnosperms	209 + 570	167 + 468	0.72	0.71	0.74	42 + 102	0.86	0.55	0.30
Monocots	870	692	0.74	0.73	0.73	178	0.83	0.11	0.12
Total	6359	5087				1272	0.85	0.57	0.51

Note: Pavement cells of five different plant clades were used for prediction. The two minority classes angiosperms and gymnosperms were merged. The multi-class SVM was trained on 80% of the total dataset, the remaining 20% were used as a test set to evaluate the resulting SVMs.