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

Table 3

The comparison of nine improved and existing feature-selection methods with respect to micro-𝐹1 measure on 20-Newsgroups for NB and SVM, respectively. The bold values indicate the best performance of the classifier when various feature-selection methods are used, respectively.

	Naïve Bayes					Support vector machines
	400	800	1200	1600	2000	400	800	1200	1600	2000
IG	49.55	57.76	62.88	65.99	68.34	58.15	62.53	65.17	67.28	68.45
IGX	49.54	57.76	62.85	65.92	68.33	59.01	63.51	66.06	67.91	69.32
CHI	67.29	73.26	75.46	76.58	77.61	73.67	76.07	76.26	76.18	75.60
CHIX	69.48	74.66	76.39	77.46	78.35	75.14	77.09	77.22	77.27	76.96
MI	23.13	33.80	51.88	55.47	60.50	27.46	38.24	54.21	57.02	60.62
MIX	27.39	39.06	53.78	59.97	63.87	29.17	41.18	56.71	61.61	64.19
DF	59.23	66.37	70.32	72.78	74.07	63.57	66.69	69.57	71.23	72.06
DFX	60.88	67.35	70.97	73.62	74.94	65.71	68.17	70.37	71.92	72.97
GINI	73.79	76.68	77.83	78.21	78.74	77.13	75.58	74.50	73.99	74.26
GINIX	74.45	77.17	78.22	78.94	79.18	77.13	76.75	75.68	74.97	75.16
DIA	15.45	22.97	25.70	28.51	30.96	33.53	42.42	46.56	50.22	52.19
DIAX	47.90	59.59	66.08	73.45	76.23	59.57	67.94	72.12	76.22	77.80
CMFS	73.05	75.68	77.13	78.57	79.17	76.91	77.50	76.80	76.20	75.75
CMFSX	74.24	76.84	78.36	79.56	80.07	77.37	77.96	77.55	77.24	77.00
OCFS	46.50	55.37	60.64	64.72	66.87	54.87	60.30	63.59	66.36	67.61
OCFSX	55.49	62.33	66.25	68.36	69.89	63.08	66.51	68.60	69.45	70.21
DFPFS	67.36	71.41	73.11	74.14	74.77	69.97	71.21	72.00	72.50	72.92
DFPFSX	56.74	61.90	64.15	64.94	65.40	62.15	65.11	66.22	66.91	67.31