How would you handle imbalanced datasets in a classification problem to ensure model robustness and accuracy?

How would you handle imbalanced datasets in a classification problem to ensure model robustness and accuracy?