What are the best practices for handling imbalanced data in Machine Learning?

Resampling techniques represent one of the best practices for handling imbalanced data using Machine Learning. Such techniques handle imbalanced scenarios either by oversampling the minority class or by undersampling the majority class to result in a balanced dataset. Other approaches consist of applying specific algorithms, such as SMOTE, which are designed to treat imbalanced data.

Other approaches toward dealing with imbalanced data would be developing algorithms that specifically contemplate imbalanced data, such as SMOTE, or adjusting model assessment to other metrics. One would probably, use precision-recall curves over pure accuracy because they can give much better readings of performance under imbalanced scenarios.