Sex between models-inductive modelling using genetic algorithms

We describe GAAF (Genetic Algorithm for the Approximation of Formulae) which is a tool that, given a database of examples, will induce a model that explains those examples. GAAF was developed by Cap Volmac, a member of the Cap Gemini Sogeti Group, as part of the ESPRIT III project PAPAGENA (6857). GAAF is primarily targeted at the financial market. As such it forms the core of the OMEGA system that 'breeds' predictive models for credit-scoring, insurance, or marketing purposes. OMEGA is a collaborative development of Cap Volmac and KiQ Business Solutions Ltd., a UK-based financial consultancy firm. We also describe vanilla genetic algorithm (GA) which is a search-process that, loosely based on Darwin's evolution principle, will breed a set of individuals called a pool to increasingly higher levels of fitness. An individual, in the GA sense, in a candidate solution represented in some suitable format (e.g. a bitstring). Its fitness is an-application specific-measure of how good a solution the individual represents. >