Genetic modeling of artificial neural nets – an application to credit evaluation

Artificial neural nets (ANN) have been applied successfully to various classification problems like credit evaluation, etc. While the ANN-specific learning strategies and algorithms are independent from the application domain and the problem instance, the configuration of an ANN is a design and decision problem which should take into account the specific problem domain (i.e. credit evaluation) and the specific instantiation (i.e. given set of indicators). In this article we show how to apply ‘Genetic modeling’ to this configuration problem, i.e. we show how suitable ANN-configurations can be constructed in an evolutionary manner using a genetic algorithm. We report on empirical results for the application to the configuration of LVQ-ANN for credit evaluation.