How to handle imbalanced data
What are the problems with high-dimensional classification, and how to address them
How to handle missing data
how to do feature selection
how to capture feature interaction