fast-bert Threshold to convert probability to binary value in multi label classification

Hi, First of all thanks for the great work. it is really useful and runs smooth. I managed to run the model for my data and here is what i have got,

I have 54 labels in total and it is a multi label classification. Now as an output, i have the probabilities for each class in my output dataset. I have two questions,

How do we actually find the threshold value to convert probability to binary value ( 0 or 1) for each class?
Can we calculate classification matrix for each class from this?

And if I my understanding is right, 0.9691836689659876 is the AUC score. Is that right?

Sep 24 '19 06:09 JeevaGanesan

Yes 0.9691836689659876 is the AUC score.

You can use 0.5 as the threshold for using 0 or 1. Thats whats the accuracy_thresh metric does by default.

Sep 24 '19 13:09 kaushaltrivedi

Thank you. What is eval_accuracy_thresh? How do I interpret that value? And is there a way to build classification matrix? Or find out TP,FP,TN,FN values?

Sep 24 '19 18:09 JeevaGanesan

Hi. Is it possible to print AUC score for each class label?

Oct 03 '19 01:10 JeevaGanesan

Could you please share Databunch and learner code

Oct 16 '19 17:10 ramakth1

Hi,

Could you please share notebook files

Oct 18 '19 07:10 ramakth1

Hi. Is it possible to print AUC score for each class label?

I also have same question and would like to know if the library can support to output the AUC of each class. Thanks.

Mar 20 '20 02:03 charlie6989

Hi. Is it possible to print AUC score for each class label?

I also have same question and would like to know if the library can support to output the AUC of each class. Thanks.

Me too. Bro. Have you solved this problem? Could you share same tips to me? A lot of THANKS in advance.

Mar 17 '21 08:03 imevanlo