表 5.

本研究所提出的方法与单期相单分类器结果比较

Comparison of the results by the proposed method and those of models with a single phase and a single classifier

Phase	Strategy		AUC	ACC	SEN	SPE
KNN: k-nearest neighbour; SVM: Support vector machines; MLP: Multilayer perceptron; LR: Logistic regression; LDA: Linear discriminant analysis; GNB: Gaussian bayes; DT: Decision tree.
EAP	Worst	SPEC+GNB	0.601	0.586	0.708	0.464
	Best	JMI+SVM	0.687	0.711	0.770	0.647
	Decision Fusion		0.703	0.685	0.773	0.595
LAP	Worst	fisher_score+DT	0.616	0.612	0.618	0.615
	Best	lap_score+KNN	0.690	0.695	0.792	0.596
	Decision Fusion		0.778	0.703	0.808	0.595
PVP	Worst	SPEC+LDA	0.631	0.673	0.668	0.595
	Best	RFS+MLP	0.707	0.711	0.774	0.644
	Decision Fusion		0.773	0.711	0.789	0.631
EP	Worst	CMIM+DT	0.612	0.612	0.630	0.593
	Best	MIFS+KNN	0.729	0.684	0.823	0.536
	Decision Fusion		0.773	0.730	0.788	0.667
The proposed method			0.828	0.766	0.877	0.648