Dear Lamis

In wikipedia (Receiver operating characteristic) I found a quote that may be of interest for you on this matter:

A Unified View of Performance Metrics: Translating Threshold Choice into Expected Classification Loss

(http://jmlr.org/papers/volume13/hernandez-orallo12a/hernandez-orallo12a.pdf)

Many thanks Guillermo , i found this too "referance threshold calculation" http://www.mecs-press.org/ijigsp/ijigsp-v6-n2/IJIGSP-V6-N2-6.pdf

i try to test this method of calculating the reference threshold

Dear Guillermo,

i read the article but i didn't find response yet on my answer

I want to know how i calculate the threshold for accepting some one or rejecting him

Dear Lamis

I do not know the details of your application of ROC. But if you have a dataset of objets well classified, you may divide it in two datasets, the first one you may use to calibrate the method of classificationand the rest to estimate how the method is performing. Then you may try each of the proposed methods for calculation of the threshold and finally adopt which has the best performance.

This problem may be worked as a discrimination problem and there is software like SAS Proc Discrim with powerfull methods of clasification, several methods of evaluating the error of clasification and to model the statistical problem behind the data.

Dear Canggih,

how did select the threshold i work in verification mode too , and i stil bugin in this step i try the method of this article but this is not very efficient, could you help me please

https://www.facebook.com/860374590656405/photos/a.860376263989571.1073741827.860374590656405/860377790656085/?type=1

Dear Canggih

yes i calculate the threshold in training step, for every person thereis 3 samples

exemple :

person 1 (S1,S2,S3) i match (S1 with S2, S1 with S3, S2 with S3) i store the mean(of these scores)

person 2 (S1,S2,S3) i do the same thing , intil person n , then i will have the a vector of threshold of n person,

but this methid is not reliable, could you tel how did you calculate the threshold?

Classification Assessment Methods

This paper introduces a detailed explanation with numerical examples many classification assessment methods or classification measures such as:

Accuracy, sensitivity, specificity, ROC curve, Precision-Recall curve, AUC score and many other metrics. In this paper, many details about the ROC curve, PR curve, and Detection Error Trade-off (DET) curve. Moreover, many details about some measures which are suitable for imbalanced data are explained.

Your comments are highly appreciated.

https://ac.els-cdn.com/S2210832718301546/1-s2.0-S2210832718301546-main.pdf?_tid=20353999-8c84-422a-8191-fd1059469272&acdnat=1536185805_03b19e159f0675a01f35512961973eb5

Presentation of the paper is here

https://www.researchgate.net/publication/327403649_Classification_assessment_methods_a_detailed_tutorial