On difficulties in identification of simple linear-bilinear time-series models using memetic algorithm

The new approach to identification of linear-bilinear time-series models has been recently proposed. It is based on separated identification of linear and bilinear parts of the model and exploits the advantages of Memetic Algorithms Therefore, simple survey tests have been performed for different sets of time-series and some difficulties have been recognized. The results of the tests and possible explanations of problem are presented on following pages.