Data mining via distributed genetic programming agents

Genetic programming is a powerful search method which can be applied to the typical data mining task of finding hidden relations in datasets. We describe the architecture of a distributed data mining system in which genetic programming agents create a large amount of structurally different models which are stored in a model database. A search engine for models that is connected to this database allows interactive exploration and analysis of models, and composition of simple models to hierarchical models. The search engine is the crucial component of the system in the sense that it supports knowledge discovery and paves the way for the goal of finding interesting hidden causal relations.