Confusion Matrix for ECOC model classification Example

5 次查看（过去 30 天）
AbioEngineer
AbioEngineer 2019-8-21
评论： AbioEngineer 2019-8-31
Running this example in the Command Window:
openExample('textanalytics/CreateSimpleTextModelForClassificationExample')
Is there any way of creating a Confusion Matrix for a cross validated version of the above ECOC model? Similar to this example:
openExample('stats/TrainECOCClassifiersUsingEnsemblesAndParallelComputingExample')
Is the problem with the type of Learner during the fitECOC step? or is it because the class labels are contained in a cell array of strings, and are non-numeric like in the second example?
the error I get when using crossval is that the first input argument must be 'mse', 'mcr', or a function handle. I am trying to do a simple classification of only 2 classes, so I'd expect a 2x2 confusion matrix. I don't think I need start a parallel pool or use a very complicated learner for my NLP text classification.

请先登录，再回答此问题。

采纳的回答

Gaurav Garg
Gaurav Garg 2019-8-29
Hi,
In my understanding, you wish to know a way to create a confusion matrix for the example textanalytics/CreateSimpleTextModelForClassificationExample.
You can use confusionmat function which compute confusion matrix for classification problems. For the above example, the following code shall give you the confusion matrix -
[cm,grp]=confusionmat(YTest,YPred)
And as far as 2x2 confusion matrix is concerned, since you have multiple classes, you are expected to receive a matrix of dimensions nxn where n is the number of classes. In the output matrix, cell (i,j) denotes the class i which has been labelled as class j.
  1 个评论
AbioEngineer
AbioEngineer 2019-8-31
Thank you. But how do I get a nicely colored matrix like the second example's cross validation model?

请先登录，再进行评论。

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Classification Ensembles 的更多信息

标签

产品


版本

R2018b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by