About Versions.txt

The algorithm has three phases: (1) Initialization; (2) Adjustment of Prototypes and (3) Insertion. The second phase
may be carried out with any search algorithm. In this research each 'main' version uses a different
search algorithm. All of them with their respective modifications will be listed below.


DCIAGSA (Dynamic Centroid Insertion and Adjustment with Gravity Search Algorithm)
 - The first algorithm implemented

	M1 : data were normalized
	M2 : data were normalized and divided with DOB-SCV

DCIASGSA (Dynamic Centroid Insertion and Adjustment with Simple Gravity Search Algorithm)
 - GSA was replaced with SGSA. Data is always normalized and train and test subsets are created
 with DOB-SCV

	M1 : Data are undersampled with Tomek-links and then oversampled with SMOTE like algorithm
	M2 : Data are oversampled with SMOTE like algorithm and then undersampled with Tomek-links
	M3 : Feature Selection with tSNE Matlab function
	M4 : Feature Selection with fscnca Matlab function
	M5 : FS with CSO and AUCarea as fitness. SGSA remains unaltered.

DCIAPSO (Dynamic Centroid Insertion and Adjustment with Particle Swarm Optimization)
 - GSA was replaced with PSO. Data is always normalized and train and test subsets are created
 with DOB-SCV

	M1 : AUCarea replaces MAUC as fitness function
	M2 : Feature Selection with CSO
	M3 : New fitness for M2 FS with CSO
		fitness : sum of centroids distances
	M4 : FS with WOA
	M5 : FS with SYMON
	M6 : M5 with different fitness
		fitness: (mean_outer - std_outer) - (mean_inner + std_inner)
			in which 
				mean/std_outter are vectors containing mean and std of distances from each inclass instances to all outclass instances
				mean/std_inner are vectors containing mean and std of inner class distances
	M7 : M2 with M6 fitness
	M8 : M7 with slightly different fitness
	M9 : M8 with slightly different fitness
	M10: M2 with AUCarea as FS fitness function
	M11: M4 with AUCarea instead of MAUC for fitness calculation

DCIACSO (Dynamic Centroid Insertion and Adjustment with Competitive Swarm Optimization)
 - PSO was replaced with CSO in DCIAPSO.M10