Hello everyone ! I add two new files to my repository: 1) CohenKappa.…

…m for evaluating the classification and 2) CV_MultiSVDD for the cross validation.

CV_MultiSVDD.m

@@ -0,0 +1,70 @@
+function [param_star, C_star, err_matrix] = ...
+    CV_MultiSVDD(X, Y, kernel,nrip, KerPar,Cpar)
+
+% Cross Validation function for MC-SVDD
+
+    Num_class = length(unique(Y));
+    minimum_abs = 100;
+
+    for rip = 1:nrip
+
+        disp(['--->', num2str(rip)])
+
+        cv = cvpartition(Y,'HoldOut',0.3, 'Stratify',true);
+        idx = cv.test;
+
+        Xtr = X(~idx,:); Ytr = Y(~idx,:);
+        Xvl = X(idx,:); Yvl = Y(idx,:);
+
+        err_matrix = zeros(size(KerPar,2),size(Cpar,1));
+
+        i = 0; 
+
+        for param = KerPar
+
+            i = i + 1;
+
+            j = 0;
+
+            for C = Cpar'
+
+                C = C';
+
+                j = j + 1;
+
+                [x_class, Ytr_class, Rsquared_class, a_class, SV_class, YSV_class]=...
+            NC_SVDD_TRAINING(Xtr, Ytr, Num_class, kernel, param, C);
+
+                y_predict = ...
+                    NC_SVDD_TEST(Xtr, Ytr_class, Num_class, x_class, Xvl, kernel, param, Rsquared_class);
+
+                n = size(Yvl,1);
+
+                err = (n-sum(Yvl == y_predict))/n;
+
+                err_matrix(i,j) = err;
+
+            end
+        end
+
+        minimum = min(min(err_matrix));
+
+        if minimum < minimum_abs
+
+            [x,y]=find(err_matrix==minimum);
+
+            param_star = KerPar(x(1));
+
+            C_star = Cpar(y(1),:);
+
+            minimum_abs = minimum;
+
+            disp(minimum_abs)
+
+        end
+
+    end
+
+disp('Done')
+
+end

CohenKappa.m

@@ -0,0 +1,26 @@
+function K = CohenKappa(X, Y, y_pred, Num_class)
+
+% Function that computes the CohenKappa index for multiclassification
+% algorithms
+
+    CM = ConfusionMatrix(Y, y_pred, Num_class);
+
+    c = sum(diag(CM));
+
+    s = size(X,1);
+
+    p = []; t = [];
+
+    for k =1:Num_class
+
+        p_k = sum(CM(:,k));
+
+        p = [p,p_k];
+
+        t_k = sum(CM(k,:));
+
+        t = [t,t_k];
+
+    end
+
+    K = ((c*s) - sum(p.*t))/(s^2 - sum(p.*t));