What does it mean when roc curves intersect at a point?

I am working with a data set and I have obtained the following roc curve:

As you can see, black and Asian ethnicity cross at one point (green and purple lines). Does this have any significance? Could any conclusion be drawn from this?

Note that I am dealing with the following datasets:

-transrisk_performance_by_race_ssa -transrisk_cdf_by_race_ssa.csv -totals.csv

In order to observe whether fairness affects profits.

Topic rstudio roc r machine-learning

Category Data Science

1
answered at

From a fairness point of view one might argue that one curve dominating another curve may be an indication of a model being potentially biased towards the class with the dominant ROC curve. However, if two ROC curves intersect it implies that none of the two dominates the other one. (Please note the direction of above stated conditionals. Also this is intentionally vaguely phrased since the conclusions to be drawn are limited.)

Another (not insightful) observation is that in such intersecting cases one cannot generally tell from the ROC curve plot which of the two has a higher ROC AUC (to phrase it differently: an ROC curve dominating another ROC curve implies a higher ROC AUC).

1
answered at

No,Intersection of ROC AUC curve does not bring in any value. ROC AUC curves helps you to understand the performance of your model. A straight diagonal represent what a random model will do in case of binary classification.

Higher the Area under the curve better the model is.

About

Geeks Mental is a community that publishes articles and tutorials about Web, Android, Data Science, new techniques and Linux security.