Data Mining Technology Application in False Text Information Recognition

Table 5

Comparison of classification effects between LIBSVM and SMO.

Classifier		Feature set	Time (s)	Accuracy (%)	Recall (%)	F-measure	ROC area	Growth rate (%)
LIB	Linear	FS1	0.01	85.3306	0.853	0.853	0.853	—
		FS2	0.02	91.9421	0.919	0.919	0.919	7.74
		FS3	0.02	93.1818	0.932	0.932	0.932	1.41
		FS4	0.03	93.8017	0.938	0.938	0.938	0.64
	Polynomial	FS1	0.01	86.3636	0.864	0.863	0.864	—
		FS2	0.02	92.1488	0.921	0.921	0.921	6.72
		FS3	0.03	93.1818	0.932	0.932	0.932	1.19
		FS4	0.02	92.9752	0.93	0.93	0.93	−0.21
SVM	RBF	FS1	0.02	86.7769	0.868	0.868	0.868	—
		FS2	0.04	92.5620	0.926	0.926	0.926	6.68
		FS3	0.04	94.0083	0.94	0.94	0.94	1.51
		FS4	0.04	94.2149	0.942	0.942	0.942	0.21
	Sigmoid	FS1	0.02	82.2314	0.822	0.819	0.822	—
		FS2	0.07	79.3388	0.793	0.785	0.793	−4.15
		FS3	0.06	78.7190	0.787	0.778	0.787	−0.89
		FS4	0.06	80.3719	0.804	0.797	0.804	2.44
SMO	RBF	FS1	0.05	85.9504	0.86	0.859	0.86	—
		FS2	0.06	93.5950	0.936	0.936	0.936	8.96
		FS3	0.06	94.6281	0.946	0.946	0.946	1.07
		FS4	0.05	95.0413	0.95	0.95	0.95	0.42
	polyKernel	FS1	0	86.5702	0.866	0.865	0.866	—
		FS2	0.02	92.9752	0.93	0.93	0.93	7.51
		FS3	0.01	93.8017	0.938	0.938	0.938	0.86
		FS4	0.02	94.2149	0.942	0.942	0.942	0.43