A VNS-EDA Algorithm-Based Feature Selection for Credit Risk Classification

Algorithm 1

VNS-EDA with the elitist population strategy.

	: the original features , population size , predominant population size , weighting parameters and , neighborhood operators one-flip and two-flip, and iterations
	: the best feature subsets
(1)	begin
(2)	generate random individuals of initial population according to the values of the symmetrical uncertainty (correlation measurement between features)
(3)	set the neighborhood operators one-flip and two-flip, designate an empty set as the elitist population
(5)	while (termination criteria not met) do
(6)	compute the fitness of each individuals according to equation (4) in
(7)
(8)	select the top promising individuals from as a predominant population and update the elitist population  with
(9)	use the neighborhood operators to generate new individuals in
(10)	build the probabilistic model from according to equation (5)
(11)	sample to generate new candidate individuals as a new generation
(12)
(13)	end while
(14)	the process ends when the termination criteria is satisfied