Table 3.

Robust genera subset from two feature selection methods in four classifiers.

	Logit Boost/FS	LogitBoost/BE	LMT/FS	LMT/BE	SVM/FS	SVM/BE	KNN/FS	KNN/BE	Mean of order
PSBM3	3	2	5	3	3	2	3	3	3
Candidatus Azobacteroides	6	10	7	8	10	122	5	60	28.5
Cetobacterium	10	19	6	25	19	31	17	154	35.125
Ralstonia	46	17	93	14	27	16	45	24	35.25
Proteus	32	3	126	15	6	27	9	78	37
Flavobacterium	33	7	98	51	44	17	49	7	38.25
Moryella	8	105	1	77	7	1	103	65	45.875
Citrobacter	11	89	20	5	88	7	135	13	46
Anaerofustis	23	6	35	73	66	26	129	36	49.25
Dickeya	18	26	27	10	171	11	28	111	50.25
Owenweeksia	52	16	95	6	8	131	68	58	54.25
Salmonella	22	69	99	61	49	59	125	77	70.125
Pediococcus	99	93	46	82	67	45	145	19	74.5
Variovorax	80	127	54	79	133	79	58	57	83.375
Leuconostoc	83	112	96	63	63	91	94	88	86.25
Marvinbryantia	106	156	118	43	80	113	78	89	97.875
Novosphingobium	51	151	121	48	90	82	116	151	101.25

We present 17 genera selected in combination of four classifiers and two feature selection method. Column represent “Classifier/feature selection method”. The figures in the table show the order of genera in selection steps. The lower number (figure) indicates the more importance for genera in terms of performance.