How does SVM behave when classes are highly imbalanced?

SVM can be biased toward majority class unless class weights or resampling are used.

In imbalanced datasets, SVM may prioritize accuracy on majority class. Using class_weight='balanced' or adjusting C per class helps shift margin to improve minority class detection.