AWSMOTE: An SVM-Based Adaptive Weighted SMOTE for Class-Imbalance Learning

Algorithm 2

CaseWeight.

Input: majority class , minority class , the vector of the variable weights , nearest neighbor , , and
	Output: the vector of each minority sample weight
(1)All and a set of resamples from are used to train SVM classifier
(2)Obtain the set of minority support vector ,
(3)for to do
(4) for to do
(5)  Compute the nearest neighbors of () and obtain ;
(6)  for to do
(7)   The element of variable of the new sample :
	Where is a random number
(8)  end for
(9)  Add the generated samples to set which is generated by and obtain new minority samples as testing set; the original and as training set are used to train SVM classifier, and the number of correct predictions is
(10) end for
(11) The additional weight of the support vector is
(12) The new weight of each support vector is
(13) Standardized ,
(14)end for
(15)The initial weight of the nonsupport vector , standardized
(16)return The vector of each minority sample weight