Exploiting Ordinal Class Structure in Multi-Class Classification: Application to Ovarian Cancer
By Burook Misganaw, Student Member, IEEE and M. Vidyasagar, Life Fellow, IEEE
Published June 30, 2015.
In multi-class machine learning problems, one needs to distinguish between nominal labels which do not have any natural ordering, and ordinal labels which are ordered. Ordinal labels are pervasive in biology, and some examples are given here. In this note, we point out the importance of making use of the order information when it is inherent to the problem. We demonstrate that algorithms that use this additional information outperform algorithms those that don’t, on a case study of assigning one of four labels to ovarian cancer patients on the basis of their time of progression-free survival (PFS). As an aside, it is also pointed out that algorithms that make use of ordering information require fewer data normalizations. This aspect is important in biological applications, where data is plagued by variations in platforms and protocols, batch effects, etc.