I am working on classification based problem using machine learning. I came across two abbreviation ROC and AUS which are said to be classification measures. What are the expansion of these and what does it mean?
A ROC curve is a Receiver Operator Curve. It is often used treats the output of your learning algorithm as a ranking--say a ranking of solutions by how likely they are to be in class 1. The Y Axis is the portion of true positives found as you move along the ranking from origin. The X axis is the portion of false positives. So it shows you the tradeoff between identifying any given portion of true positives and the cost of that identification in false positives.
A particularly nice feature of the ROC curve is that a line tangent to the curve identifies the point on the curve where the tradeoff between different types of errors is optimal given the slope of the tangent line represents your desired tradeoff.
As to "AUS", are you sure you don't mean "AUC?" AUC refers to the area under the ROC curve. The bigger the AUC, the better the ROC curve has performed. I don't know the "AUS" term off the top of my head.
I have used ROC curves as the fitness objective for Genetic Program many times and it works very nicely for binary classification problems. It is especially effective when there are very few examples of one of the two classes.
- Badges
- Science topic
- Similar topics
- Psychometrics
- Measurement