Automatic credit approval is important for the efficient processing of credit applications. Eight different genetic programming (GP) approaches for the classification rule mining of a credit card application dataset are investigated, using both a Booleanizing technique and strongly- typed GP. In addition, the use of GP for missing value handling is evaluated. Overall, on the Australian Credit Approval dataset, those GP approaches that had poorer classification correctness on the training data often proved better at generalizing for the test set.
|Title of host publication
|2007 IEEE Congress on Evolutionary Computation
|Published - Sept 2007