Knee Joint Vibration Signal Analysis with Matching Pursuit Decomposition and Dynamic Weighted Classifier Fusion

Table 1

Classification results of different classifiers and ensemble systems on the testing data set. LS-SVM/LOO: the least-squares support vector machine evaluated with the leave-one-out (LOO) method. CSVM: the component least-squares support vector machines. DWF: the dynamic weighted fusion method. ROC curve: receiver operating characteristic curve.

Classifier	Accuracy; (%)	Area under; ROC curve; (𝐴𝑧)	Standard error (SE)
LS-SVM/LOO	87.64	0.7523	0.0536
CSVM1	76.41	0.8578	0.0403
CSVM2	83.15	0.8772	0.0413
CSVM3	76.41	0.7884	0.0541
CSVM4	76.41	0.7064	0.0643
CSVM5	82.02	0.8844	0.0359
Bagging	78.65	0.8483	0.0474
DWF	88.76	0.9515	0.0244