Optimising the Selection of Input Variables to Increase the Predicting Accuracy of Shear Strength for Deep Beams
Algorithm 1
Selection of the best predictors by GA and mutual information.
Input: I and I
n = (1, …, m): statistical data determined by the previous algorithm (Figure 4);
: the desired number of predictors;
A: selection pressure;
: maximum number of generations;
: size of the population; and
Output: {j} set the indexes of the selected predictors.
(i)
Generate a set of chromosome for the initial population. Each chromosome is a vector = containing the indices of neuron j generated randomly without repeating elements.
(ii)
For generation = 1: ,do
(iii)
Evaluate the population.
(iv)
For idx = 1: ,do
(v)
Calculate for each using the following formula by the calculated values of mutual information for all elements of chromosomes .
(vi)
: storing the fitness of each idx;
(vii)
end for (loop).
(viii)
Rank the individuals according to their fitness .
(ix)
Store the genes of the best individual into {j}.
(x)
Perform the crossover
(xi)
.
(xii)
For idx = 1: ,do
(xiii)
.
(xiv)
Choose the indices of the parents randomly using the asymmetric distribution [60].