Accurate High Performance Concrete Prediction with an Alignment-Based Genetic Programming System

International Journal of Concrete Structures and Materials

Table 4 Parameter tuning.

Technique	Parameter name	Values tested [min;max;# of values tested]	Best value
LIN	ridge	[1.0E−7;1.0E−9;3]	1.00E−08
LIN	eliminateColinearAttributes	True; False	True
ISO	–	–	–
K*	globalBlend	[0;100;10]	30
MLP	learningRate	[0.1;0.4;4]	0.15
	momentum	[0.1;0.4;4]	0.1
	hiddenLayers	[1, 7]	3
	trainingTime	[500;1000;5]	1000
RBF	minStdDev	[0.1;0.5;5]	0.2
RBF	ridge	[1.0E−7;1.0E−9;3]	1.00E−08
SVM	DegreePolynomialKernel	[1, 4]	2
SVM	regOptimizer	RegSMO; RegSMOimproved	RegSMOimproved

For each technique, the table reports the tuned parameters and the value used in the experiments that were performed. The reader is referred to the Weka ML tool documentation (Weka 2018) for the explanation of these parameters.