Binary Bitwise Artificial Bee Colony as Feature Selection Optimization Approach within Taguchi’s T-Method

Table 5

The prediction accuracy (MAE) compilation across all proposed enhanced methods and existing Taguchi’s T-Method on validation datasets for large sample data.

Dataset	# sample training	# sample testing	# features	Measure	T-Method [28]	T Method-BitABC and T Method-PBPSO [26]	T-Method + OA [28]	% MAE improvement of best result vs. T-Method
Body fat	176	76	14	MAE	0.3868	1.5663	1.651	NA
				SD	0.3250	1.0743	1.056
				# optimum features	14	10	9
Abalone	2924	1253	8	MAE	4.2377	3.6447	3.757	13.99%
				SD	3.3917	2.5686	2.791
				# optimum features	8	3	4
Heating	538	230	8	MAE	8.6033	5.7762	6.317	32.86%
				SD	8.5813	3.5191	3.831
				# optimum features	8	5	4
Cooling	538	230	8	MAE	8.1106	5.9515	6.312	26.62%
				SD	4.5612	4.0778	4.217
				# optimum features	8	5	2
Concrete Compressive Strength	721	309	7	MAE	11.4115	11.8847	12.382	NA
				SD	11.3301	12.3216	11.677
				# optimum features	7	6	6
Auto MPG	274	118	7	MAE	6.0035	3.5716	3.259	45.71%
				SD	3.0157	2.6609	2.717
				# optimum features	7	2	5