Class Balancing refers to techniques aiming to resolve the biases in classification models that occur due to imbalance in the number of samples per class. For example: if, in a two-class problem, Class 1 has 900 samples and Class 2 has only 100, a model that classifies any sample as from Class 1 still has good global accuracy. One simple class balancing technique is to undersample from the classes with more elements. Another is to oversample (as in 'sample with replacement') from the classes with less elements. But there are many other approaches, such as data augmentation, and adjusting the 'class weights' during training.