Table 2.

SMOTEC-SMO algorithm descriptions.

Input: Number of minority class samples N; the number of clusters K; similarity threshold S; SMOTE oversampling expansion multiplier T%.

Output: SMO classification plane and prediction results.

Initialize the hypermastigote, which include the number of iterations t, the learning rate L, the hypermastigote of the recurrent neural network β1, β2, and the computational gradient of the model gt;

(1) BEGIN/∗ Improved SMOTE Over-sampling∗/

(2) G = int(N ∗ T%); //Oversampling a few class examples

(3) For i = 1 to G

(4) For I = 1 to N

(5) Compute k-NN xi; //Find 5 nearest neighbors of xi

(6) For i = 1 to nuberattr/∗nuberattrIndicates the number of attributes∗/

(7) If xij is a numeric attribute

(8) If if xij ∈  Minority then xnj=(xi − xki)∗

(9) Rand(1)//Different generation methods for different cases of near neighbors

(10) elseif if xij ∈  Minority then xnj=(xi − xki)∗

(11) ∗rand(0.5)

(12) endif

(13) endif

(14) endFor

(15) endFor

(16) endFor

(17) k − NN(xj, xj′)

(18) If j ∈ Minority then xnj=(xi − xki)∗, /∗ If j is a continuous value ∗/

(19) If j  ∈  nominal then δn∑j=1c|Nj, xj, c/Nj, x − Nj, xj′, c/Nj, x′|(VDM) If J is a discrete value Δ=δc+δn / ∗∗∗ Clustering under-samping  ∗∗∗/

(20) Clustering divides the training data set into K clusters;

/ ∗∗∗∗∗∗∗∗ SMO  ∗∗∗∗∗∗/SMOTraining; //Training with sequential minimum optimization (SMO) algorithm

SMOProjections;

END