Ordered incremental training for GA-based classifiers

This paper proposes ordered incremental genetic algorithms (OIGAs) to address the incremental training of input attributes for GA-based classifiers. Rather than learning attributes in batch as with normal GAs, OIGAs learn attributes one after another. Attributes are first arranged in different orders by evaluating their individual discriminating ability. Classification rule sets are then evolved incrementally to accommodate attributes continuously. By experimenting with different attribute orders, different approaches of OIGAs are evaluated using benchmark classification data sets. The simulation results show that OIGAs can achieve generally better performance than normal GAs, and OIGAs perform the best when training with a descending order of attributes.