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

Table 8

The comparison of nine improved and existing feature-selection methods with respect to AUC 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	0.8051	0.8253	0.8369	0.8399	0.8442	0.8933	0.9044	0.9093	0.9128	0.9120
IGX	0.8055	0.8242	0.8357	0.8398	0.8444	0.8973	0.9080	0.9134	0.9153	0.9151
CHI	0.8214	0.8312	0.8365	0.8402	0.8423	0.9090	0.9114	0.9095	0.9131	0.9133
CHIX	0.8345	0.8416	0.8478	0.8521	0.8520	0.9130	0.9137	0.9158	0.9191	0.9170
MI	0.5009	0.5030	0.5113	0.5419	0.6082	0.6357	0.6618	0.7137	0.7478	0.7620
MIX	0.5237	0.5505	0.5925	0.6359	0.6593	0.5681	0.6359	0.6653	0.7024	0.7310
DF	0.8041	0.8242	0.8344	0.8404	0.8422	0.8952	0.9078	0.9157	0.9140	0.9143
DFX	0.8051	0.8262	0.8382	0.8432	0.8439	0.8981	0.9103	0.9181	0.9174	0.9168
GINI	0.8177	0.8358	0.8463	0.8476	0.8492	0.9105	0.9144	0.9171	0.9155	0.9138
GINIX	0.8229	0.8420	0.8480	0.8534	0.8533	0.9123	0.9179	0.9229	0.9184	0.9174
DIA	0.6252	0.6372	0.6058	0.6192	0.6465	0.7867	0.8151	0.8420	0.8575	0.8696
DIAX	0.5700	0.5931	0.6111	0.6279	0.6511	0.5791	0.6142	0.6473	0.6842	0.7052
CMFS	0.8132	0.8216	0.8354	0.8397	0.8430	0.9067	0.9098	0.9123	0.9181	0.9152
CMFSX	0.8183	0.8351	0.8445	0.8475	0.8513	0.9079	0.9123	0.9165	0.9185	0.9190
OCFS	0.8189	0.8349	0.8436	0.8453	0.8490	0.9106	0.9041	0.9119	0.9160	0.9158
OCFSX	0.8244	0.8401	0.8480	0.8488	0.8508	0.9151	0.9112	0.9141	0.9168	0.9191
DFPFS	0.7504	0.7507	0.7511	0.7517	0.7525	0.8735	0.8750	0.8743	0.8741	0.8729
DFPFSX	0.7723	0.7770	0.7772	0.7782	0.7786	0.8815	0.8824	0.8819	0.8806	0.8809