Outputs of classifiers and how to measure confidence?

Question

Jack Latham 2017-12-19

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/373669-outputs-of-classifiers-and-how-to-measure-confidence

评论： Jack Latham 2017-12-20

exampleData.mat

Hi, so I'm using the classification learner to train various models. I've found that the trained models can output (as well the predictions yfit) 2 other variables; I can't find any info regarding these in the documentation - can anyone shed any light? the first appears to weights, where the largest value corresponds to the predicted label. I've attached a mat file containing an example of these outputs, as well as the corresponding correct labels.

I would like to be able to calculate the confidence of a prediction - can anyone suggest a means of doing this?

Many thanks

2 个评论
显示无隐藏无

Image Analyst 2017-12-19

Try attaching your data. http://www.mathworks.com/matlabcentral/answers/6200-tutorial-how-to-ask-a-question-on-answers-and-get-a-fast-answer

Jack Latham 2017-12-20

Attached, should make it clearer

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Bernhard Suhm 2017-12-20

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/373669-outputs-of-classifiers-and-how-to-measure-confidence#answer_297072

You could try using the predint or confint functions, see also https://www.mathworks.com/matlabcentral/answers/266432-how-to-plot-prediction-and-confidence-interval