Research on Hybrid Feature Selection Method Based on Iterative Approximation Markov Blanket

Table 4

Experimental results of five datasets of filter-independent features.

	WYHXB		NYWZ		BlogData		RBuild		CCrime
	Number of features	Ave-RMSE	Number of features	Ave-RMSE	Number of features	Ave-RMSE	Number of features	Ave-RMSE	Number of features	Ave-RMSE
0.95	758	234.960328	9768	233.324863	266	12.645784	97	354.101779	120	0.131535
0.9	718	235.019819	9254	233.324863	252	12.645784	92	354.090179	114	0.131695
0.85	678	234.800187	8740	233.324863	238	12.645784	87	354.134252	107	0.131858
0.8	638	235.133101	8226	233.324863	224	12.645784	82	354.146541	101	0.131792
0.75	598	235.104648	7712	233.388367	210	12.645784	77	353.914801	95	0.131897
0.7	558	235.132128	7198	233.388367	196	12.645784	72	353.914801	88	0.131853
0.65	518	235.191663	6683	233.385479	182	12.645784	66	353.923275	82	0.131902
0.6	478	235.202756	6169	233.394604	168	12.645784	61	354.042364	76	0.132113
0.55	438	235.263138	5655	233.394604	154	12.645784	56	354.050328	69	0.132164
0.5	399	235.962421	5141	233.357302	140	12.645784	51	354.053246	63	0.132310
0.45	359	235.941428	4627	233.355757	126	12.649723	46	354.770411	57	0.132497
0.4	319	236.399412	4113	233.354086	112	12.651157	41	354.849084	50	0.132620
0.35	279	236.574098	3599	233.354248	98	12.657242	36	355.659524	44	0.133428
0.3	239	376.546789	3084	233.358374	84	12.664293	30	355.714190	38	0.133759
0.25	199	406.768586	2570	233.399275	70	12.671595	25	355.700106	31	0.134865
0.2	159	445.621765	2056	233.437486	56	12.676944	20	355.714027	25	0.136386
0.15	119	545.521345	1542	233.539485	42	12.677181	15	355.722452	19	0.137433
0.1	79	553.326100	1028	233.550540	28	12.677343	10	355.785519	12	0.139937