How do you handle imbalanced datasets in machine learning?

I’m working with imbalanced datasets in machine learning. What techniques, like oversampling, undersampling, and using different evaluation metrics, can I apply to address the imbalance and improve the performance of predictive models?