Open Internet by MindsNet

Misapplication of Class Imbalance Solutions

The common approach to addressing class imbalance in machine learning is misguided. Class imbalance is not a problem in itself but becomes one when the wrong metric is optimized, training objectives assume balanced priors, or business costs are asymmetric.

Computing & Technology, Computer Science, Machine Learning