Hello, I hope everyone is doing fine. I am looking for help regarding ROC Curve. Actually, I have a confusion matrix with all other statistics like accuracy, specificity, sensitivity, MCC coefficient, etc. I am trying to plot the ROC curve now. Is this possible using the confusion matrix and this information?
If it is possible, please guide me through the process.
Thank You.
There are many resources in the internet about this. You can have a look at youtube videos. My 2 cents, a confusion matrix does not suffice. You need to have many such matrices to form a ROC curve, actually each point the ROC curve needs such a matrix.
Dear Ahmed,
ROC curve can be applied when you have two vectors : one corresponds to the predictive variable in y, and the other to the groups (only two groups, coded for example -1 and +1).
You will try to find the best threshold, applied to the predictive variables, which separates at the best the to groups. For this purpose, you will iteratively test a threshold, from the smallest value in y to the highest one. At each tested threshold, you can attach a confusion matrix. Thus, you will have many confusion matrices. From these confusion matrices, you willl extract two numbers: the proportion of rigthly classified observations in the +1 group and the proportion of wrongly classified observations in the -1 group.
The ROC curve is the representation of these two values, for each threshold.
You thus need a lot of confusion matrices .
It is easy to find more details on the Web, simply typing "roc curve explained".
Kind regards,
Dominique
https://www.researchgate.net/post/What_should_I_do_to_draw_a_ROC_curve_for_two_countinous_variable_in_different_directions_and_also_combine_them_for_check_additive_predictive_value - Receiver operating characteristics .
Hello, thank you very much Mehmet Sinan Iyisoy Dominique Bertrand Jaydip Datta for your quick and valuable responses to my question.
- Badges
- Science topic
- Similar topics
- Counseling Psychology
- Confusion