A Computer-Aided Diagnosis System Using Deep Learning for Multiclass Skin Lesion Classification

Table 2

Classification accuracy of fine-tuned ResNet-101 deep features using augmented HAM10000 dataset.

Classifier	Recall rate (%)	Precision rate (%)	FNR (%)	AUC	Accuracy (%)	Time (sec)	F1-score (%)
LSVM	86.00	86.28	14.00	0.98	85.5	746.2	86.14
QSVM	91.57	91.71	8.428	0.992	91.1	1010.2	91.64
CSVM	92.71	92.42	7.285	0.992	92.1	11321.1	92.56
MGSVM	89.42	89.42	10.57	0.988	88.9	1919.4	89.42
CKNN	78.87	77.85	21.14	0.961	77.2	268.2	78.36
CKNN	58.42	63.00	41.27	0.887	58.9	263.6	60.62
WKNN	84.85	84.00	15.14	0.977	83.3	260.5	84.42
EBT	56.57	56.57	43.42	0.855	57.2	1544.5	56.57
ESKNN	80.28	92.28	19.71	0.99	92.1	4590.9	85.86
ESD	85.85	86.28	14.14	0.98	85.6	821.79	86.06

The bold value represents best ones.