Evaluation of Feature Selection Methods for Mammographic Breast Cancer Diagnosis in a Unified Framework

Table 4

Feature selection results. The top-most important features that achieve AUC larger than 0.85 are in bold to each FS method.

	The most to the least important features
CFS	16	7	14	3	11	5	15	6	2	8	13	17	10	9	1	4	12	18
ECFS	8	9	17	4	10	2	1	16	12	3	14	6	13	15	7	11	5	18
ILFS	11	14	18	5	3	15	13	1	4	2	10	6	9	7	16	12	8	17
LAPLACIAN	8	5	4	3	9	2	1	16	7	18	6	11	15	10	13	14	17	12
LASSO	17	18	15	13	6	16	4	1	2	8	9	5	3	7	11	14	10	12
LLCFS	3	5	4	2	1	8	9	7	16	11	18	6	15	10	14	13	17	12
RELIEFF	10	14	11	7	18	8	4	12	3	9	13	17	16	6	15	5	1	2
ROC	9	17	4	8	10	2	1	16	3	12	11	15	6	14	13	18	7	5
UFSOL	9	1	2	3	5	4	8	16	7	11	18	6	17	12	15	10	13	14
WILCOXON	10	16	9	17	4	12	6	8	14	2	1	13	3	18	7	11	15	5