A0651
Title: Proper accuracy metrics in multi class classification when sub-classes are involved
Authors: Lili Tian - State University of New York (United States) [presenting]
Abstract: The importance of using proper classification metrics in biomarker evaluation, as well as other classification tasks, has been severely overlooked in scientific research. The purposes are threefold: 1) to review some key concepts, including classification types and classification metrics in multi-class classification; 2) to address some common pitfalls in multi-class classification caused by using improper metrics and highlight the importance of matching classification metrics to the research aims; 3) to present some recently developed metrics for multi-class classification when sub-classes are involved, as well as related statistical inference methods.
