1
$\begingroup$

I am working on some clustering algorithms like DBSCAN and local outlier factor. Now i want to know how can I make ROC and AUC curves from clustering results. Do anyone know how can i make RO and AUC curves. Thank You

$\endgroup$
7
  • $\begingroup$ in which programming language $\endgroup$
    – develarist
    Commented Sep 25, 2020 at 5:55
  • $\begingroup$ I am using python $\endgroup$ Commented Sep 25, 2020 at 5:58
  • $\begingroup$ What would a ROC curve mean in this context? A ROC curve plots changing rates of true & false positives as a threshold is manipulated. Do you have knowledge of the true state of each point outside of the clustering? (If so, why are you using clustering instead of some method that incorporates that information?) What constitutes the threshold that can vary in your case? Moreover, what do you want a ROC curve for? $\endgroup$ Commented Sep 25, 2020 at 6:53
  • $\begingroup$ ROC curves are defined for binary classification problems, not for clustering. Maybe you can take a step back and explain what you want to achieve? $\endgroup$
    – Calimo
    Commented Sep 25, 2020 at 8:30
  • $\begingroup$ @gung-ReinstateMonica I do not have true information so I might not be able to get ROC curve. $\endgroup$ Commented Sep 25, 2020 at 9:38

0