Calculate Sensitivity and Specificity from Code generated from Classification Learner

I have trained my dataset in the classification learner app and tried to calculate classification performance using leave-one-out cross-validation. Since classification learner doesn't support this configuration of K-fold, I used the way of generating the code for training the currently selected model.

I have tried to compute the sensitivity and specificity but all the ways I found depend on predicted class labels and I can't get the resulted class labels since it is not a new dataset. I just want to evaluate the trained model.

Is any way to evaluate the sensitivity and specifity or the confusion matrix from Classification Learner App Code generated?

Answer 1

[Sarah Ayyad]

I computed all performance metrics by the following way

[validationPredictions, validationScores] = kfoldPredict(partitionedModel);
confmat = confusionmat(response,validationPredictions) % where response is the last column in the dataset representing a class
TP = confmat(2, 2);
TN = confmat(1, 1);
FP = confmat(1, 2);
FN = confmat(2, 1);
Accuracy = (TP + TN) / (TP + TN + FP + FN);
Sensitivity = TP / (FN + TP);
specificity = TN / (TN + FP);

z = FP / (FP+TN);
X = [0;Sensitivity;1];
Y = [0;z;1];

AUC = trapz(Y,X);  % This way is used for only binary classification

[Raghdah Wadi]

sarah ,can you contact me please

[e-mail]

@Sarah Ayyad

[Shen Jia Wei]

Thanks a lot !!!!!!!!!!!!!!!!!