Symbolic regression by genetic programming: using the prediction of PCDDs/PCDFs emissions from incinerators as an example

摘要

Recent advances of genetic algorithm have resulted in a dramatic grwoth of optimization and prediction techniques for Inonlinear models. The use of a tree-structured genetic algorithm for nonlinear model identification can be regarded as an evolutonary technique in system identification. Such a genetic programming approach is particularly adapted in this paper for the system identification of nonlinear structure with the unique condition of small scale samples. Example is drawn from the emission test of PCDDs/PCDFs through the flue gas discharge from several municipal incineators in Canada. It shows genetic programming may successfully solves the representation problem of nonlinear models with higher flexibility while considering the inherent problem-oriented mechanism existing in real world systems.