Improved Feature-Selection Method Considering the Imbalance Problem in Text Categorization

Table 7

The comparison of nine improved and existing feature-selection methods with respect to micro-𝐹1 measure on WebKB for NB and SVM, respectively. The bold values indicate the best performance of the classifier when various feature-selection methods are used, respectively.

Feature selection	Naïve Bayes					Support vector machines
	400	800	1200	1600	2000	400	800	1200	1600	2000
IG	71.36	74.17	75.84	76.61	77.46	83.52	85.66	86.10	86.80	86.77
IGX	71.64	73.96	75.64	76.57	77.48	84.10	86.29	86.76	87.27	87.33
CHI	71.92	74.16	74.65	76.19	76.93	85.88	86.52	86.08	86.89	86.89
CHIX	74.98	76.28	77.39	78.57	78.83	86.69	86.97	87.26	87.63	87.34
MI	33.96	34.63	36.32	40.71	49.58	44.96	48.60	57.23	61.75	64.44
MIX	33.69	38.31	44.49	51.00	54.30	38.37	46.86	50.56	56.63	60.85
DF	71.04	73.74	75.82	76.87	77.23	83.84	85.87	87.27	87.04	87.05
DFX	71.93	74.48	76.57	77.52	77.55	84.51	86.28	87.73	87.64	87.59
GINI	72.13	75.58	77.68	77.84	78.42	86.17	86.81	87.09	87.13	86.90
GINIX	73.75	76.96	78.06	78.80	78.97	86.61	87.52	88.20	87.56	87.44
DIA	47.75	48.40	46.06	48.77	51.00	62.66	69.13	74.84	76.95	79.03
DIAX	50.11	53.23	60.99	63.86	63.68	56.42	63.42	69.97	71.73	71.36
CMFS	72.14	73.22	75.62	76.81	77.33	85.64	86.06	86.72	87.63	87.17
CMFSX	73.37	75.89	77.63	78.06	78.70	86.04	86.66	87.29	87.54	87.75
OCFS	73.07	75.23	76.46	77.09	78.02	84.39	85.64	86.49	86.84	87.04
OCFSX	73.79	76.52	78.05	78.27	78.53	87.19	86.86	86.85	87.44	87.78
DFPFS	65.17	64.74	64.65	64.66	65.33	80.69	81.46	81.27	81.26	80.28
DFPFSX	67.96	68.81	68.37	68.58	68.83	81.83	82.08	82.11	82.00	82.01