Abstract:

Model building with observed data has two steps, which first determines model structure and is most hard, and then estimate parameters. Evolutionary modeling based on genetic algorithms can obtain a set of optimized models by simulating natural selection and evolutionary of organism. It will create a set
of model expressions at random, and adjust their structures with eproduction, crossover and mutation operation, and estimate parameters simultaneity by nonlinear regression. The explicit simple models can be derived from evolutionary modeling and have superiority interpolation and extrapolation to artificial neural network.

Key words: evolutionary computation, genetic algorithms, adaptive modeling, structure, optimization