Splits the data into categories in such a way that maximizes the information gain