How would you perform accurate classification on a very large dataset by just looking at a sample of it
