Plot confusion doesn't work

Question

Olga Zhukova 2013-12-17

3
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/109958-plot-confusion-doesn-t-work

评论： Tapan 2023-8-11

Hello,

I have the problem that when I try use plotconfusion, this function doesn't work.

I have dataset with 15 classes and I try to predict the target value using knn-classification. I've divided datasets to training and test datasets (75:25 accordinaly). My dataset has 300 instances and 90 attributes.

The problem is that when I try to call this plotconfusion function I just see that this doesn't work (it somehow just go to a infinite cycle or something like this, the process doesn't terminate). Could you tell me what's the problem or do I use it wrong?

Here the part of my code: knn = ClassificationKNN.fit(XtrainNN,YtrainNN,'NumNeighbors',5); Y_knn = knn.predict(XtestNN); loss(knn, XtestNN, YtestNN) plotconfusion(Y_knn,YtestNN)

3 个评论
显示 1更早的评论隐藏 1更早的评论

Baran Yildiz 2017-9-26

I am also having the same problem. Plot confusion doesn't seem to work even for the sample problem/dataset given in the reference link below:

https://au.mathworks.com/help/nnet/ref/plotconfusion.html#inputarg_targets

Tapan 2023-8-11

My error is please tell me how to solve

Error using plotconfusion>standard_args (line 255) Value is not a matrix or cell array.

Error in plotconfusion (line 111) update_args = standard_args(args{:});

Error in dltt (line 18) plotconfusion(testdata.Labels, Predicted);

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Nathan DeJong 2017-9-27

6
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/109958-plot-confusion-doesn-t-work#answer_283358

Try transposing the inputs so that they are row vectors rather than column vectors. It worked for me. Seems to be a strange bug in plotconfusion().

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Pedro Borges 2018-10-17

transposing the inputs worked for me, too! thanks!

请先登录，再进行评论。

Answer 2

Hamid Salimi 2021-6-9

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/109958-plot-confusion-doesn-t-work#answer_721165

I write it for anyone that may have the same problem, I solved it by converting my actual and predicted results to categorical data! your actual and predicted should be n * 1, and then use it:

plotconfusion(categorical(actual),categorical(predicted));

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。

Answer 3

Ilya 2013-12-17

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/109958-plot-confusion-doesn-t-work#answer_118669

在 MATLAB Online 中打开

I never used plotconfusion, but you can get what you want using functions confusionmat and imagesc. For example,

knn = ClassificationKNN.fit(XtrainNN,YtrainNN,'NumNeighbors',5); 
Y_knn = knn.predict(XtestNN); 
cm = confusionmat(YtestNN,Y_knn);
imagesc(cm);
colorbar;