Plot Confusion Matrix for a Trained Classifier

Question

Alessandro Fascetti 2018-10-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/422262-plot-confusion-matrix-for-a-trained-classifier

评论： Towfeeq Fairooz 2021-12-19

I trained a classifier for 7500 instances and 3 classes. Since the confusion matrix tab inside the Classifier App will not let me change font size and title (the most absurd thing ever...) I had to export the classifier as a function and do it manually. So I calculate the validationPredictions as suggested in the generated .m file

partitionedModel = crossval(trainedClassifier.ClassificationSVM, 'KFold', 5);
% Compute validation accuracy
validationAccuracy = 1 - kfoldLoss(partitionedModel, 'LossFun', 'ClassifError');
% Compute validation predictions and scores
[validationPredictions, validationScores] = kfoldPredict(partitionedModel);

This should be trivial at this point, but apparently it is not. When using:

C = confusionmat(response, validationPredictions);

the output C matrix is a 3-by-3 containing the real vs predicted classes. Then running:

plotconfusion(response,validationPredictions)

hangs Matlab and nothing shows up (this should take a split second). What is the issue?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

James Herman 2019-7-11

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/422262-plot-confusion-matrix-for-a-trained-classifier#answer_382780

Alessandro - I figured out why this isn't working for me, the data type of the arguments passed to the function (plotconfusion) needs to be "categorical". Even if you're passing a list of integers you need to convert it using the "categorical" function.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Towfeeq Fairooz 2021-12-19

It has been a long time but I thought this might help someone. So.

You are correct; I experienced the same problem and was only able to resolve it by adding categorical to the arguments of classifier app's (export function)

However, I did not clip any sections of the trainedClassier code it remains well within the code.

inputData =[X,y]; % X = data features, y = target

[myClassifier, myValidationAccuracy] = trainClassifier(inputData);

y_predicted = myClassifier.predictFcn(X_inputData);

plotconfusion(categorical(y_inputData),categorical(y predicted));

请先登录，再进行评论。

Answer 2

Ali Moshrefzadeh 2019-5-26

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/422262-plot-confusion-matrix-for-a-trained-classifier#answer_376667

The form of the input matrixes for

plotconfusion(targets,outputs)

is different form the output of

C = confusionmat

You should either use confusionchart(C) to plot the output of confusionmat as a confusion matrix or prepare the proper matrixes for plotconfusion.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

James Herman 2019-7-10

You're incorrect. Alessandro Fascetti has used the function precisedly as it is intended. I have the same issue - the function "plotconfusion" doesn't return an error, it simply hangs up during execution.

请先登录，再进行评论。