Satellite Fault Diagnosis Using Support Vector Machines Based on a Hybrid Voting Mechanism

Table 3

The FRR and FAR of HVM-SVM compared with SVM, NN, and KNN (only list parts of the results).

A total of 10000 records		5000 for training; 5000 for testing	2000 for training; 8000 for testing	1000 for training; 9000 for testing	500 for training; 9500 for testing	100 for training; 9900 for testing
SVM	FRR (fault )	0.9990301	0.9981768	0.9902465	0.9897119	0.006655164
	FAR ()	0.00048497	0.00060772	0.005689515	0.00617284	0.9933448
	FAR ()	0.0004849	0.00121544	0.004063939	0.004115226	0
HVM-SVM	FRR (fault )	1	0.9990884	0.9929558	0.9927984	0.6958344
	FAR ()	0	0.00091158	0.00270929	0.00308642	0.3041653
	FAR ()	0	0	0.00379301	0.00360082	0
NN	FRR (fault )	0.8962306	0.9541473	0.9788487	0.9405237	0.6340591
	FAR ()	0.1037694	0.04585265	0.0111413	0.01945227	0.1875932
	FAR ()	0	0	0.01001	0.040024	0.1783477
KNN	FRR (fault )	0.9960396	0.9792714	0.9149528	0.8333333	0.4125697
	FAR ()	0.0039603	0.00188442	0.00166759	0.00157729	0.204764
	FAR ()	0	0.01884422	0.0794886	0.139327	0.382666