About AUC

The aim of this web is to popularize the effective computation of AUC based on the Bubble sorting approach - the BSA. Just look at fhe following graphs comparing the implementation based on the convetinoal interpretation, the optimized version and the BSA approach. If you want to compute exact value of multi-class AUC, maybe we have an effective solution for you. Look at the graphs below.

What are the advantages of AUC interpreted by BSA?
- very comprihensible interpretation (next to the graphical)
- you can compute AUC directly from the data without any computer-aided or graphical software
- as far as we know (2011) the fastest algorithm for computation of exact value of multi-class AUC

What you can find here:
- explanation of the BSA principle
- simple examples explaning the principle
- downloads of source codes and libraries in Matlab and C/C++
- links to the papers about AUC by BSA

What you will not find here:
- explanation of conventional interpretations (probabilistic, graphical)
- relations with other similar measures (Gini Index, Sommers d, Mann-Whitney test, ...)
- overview of methods approaximating AUC
- judgement for the use of AUC or other related topics

If you have any comments or questions, don¨t hesitate to contact us:
honzikp@feec.vutbr.cz - BSA method, Matlab solution
kucera@feec.vutbr.cz - C/C++ and dll