Department of Statistics
So, where’s the profit in DM?
Automated modeling has become very powerful, avoiding problems of over-fitting.
Role for expert judgment remains
What data to use?
Which features to try first?
What are the economics of the prediction errors?