Machine Learning-Based Technique for the Severity Classification of Sublingual Varices according to Traditional Chinese Medicine

Table 1

A comparison of different ML methods by using features extracted with PCA + SIR or CNN methods as well as the original images. The numbers in the table are the mean accuracies in %, and The number in square brackets is the 95% CI. Training time measured in seconds.

MLs	ML-original (%)	ML-PCA (%)	ML-PCA + SIR (%)	ML-CNN (%)	Time (s)
SVM (RBF)	78.0 [74.7-81.3]	79.0 [76.3-81.7]	83.8 [81.0-86.5]	80.7 [77.5-84.0]	11.7
SVM (linear)	84.5 [81.6-87.4]	83.3 [79.1-87.4]	83.5 [79.8-87.2]	86.2 [82.8-89.7]	40.2
Linear model	84.8 [82.0-87.5]	84.5 [81.5-87.5]	84.5 [81.5-87.5]	83.5 [79.5-87.5]	2.2
Ridge classifier	84.7 [82.3-87.2]	85.0 [81.9-88.1]	84.5 [81.5-87.5]	87.5 [83.8-91.2]	1.6
KNN	76.2 [72.5-80.0]	76.8 [73.2-80.3]	84.0 [80.9-87.1]	82.0 [79.0-85.0]	8.9
Decision tree	67.7 [61.6-73.9]	60.5 [55.8-65.2)	83.8 [80.3-87.2)	77.8 [74.7-80.8]	14.4
Random forest	75.5 [71.3-79.7]	73.2 [69.3-77.2]	83.8 [80.3-87.2]	83.8 [81.1-86.4]	21.4