Increasing Minority Recall Support Vector Machine Model for Imbalanced Data Classification

Algorithm 1

Algorithm to increase the recall rate to 1.

	Algorithm:
	Positive margin adjustment using SVM.
	Given: a sequence of N examples XTrain and XValidation
	Output: G #Output combination classifier
	Variables:
	#Karush–Kuhn–Tucker conditions (KKT) initial alpha
	#G-means value
	Cp, Cn, k #Positive cost, Negative cost, Positive margin calibration variable
	T #the selected running iterations
	Function:
	S #classifier model
	RG-means (G) #Obtain the Recall G-means values from G
	Begin
	Initialize
	= 0
	= 0
	T = 1
	Set the 3D grid search range of Cp, Cn, k:
(a)	Select optimization variables and and solve the optimization problem using the sequential minimal optimization (SMO) algorithm to obtain and , and update to .
(b)	If the KKT conditions in (16)–(19) are satisfied within the allowable range of precision, , the KKT condition can be used for the next step; otherwise, continue with process (b).
(c)	Get .
(d)	Finally, is obtained, and and are calculated as follows:
	nstruct a classifier model S =
(e)	G = sign (S)
(f)	-means (G(XValidation)) at a condition recall of 1
(g)	If , then
	Return Gt
	End