A robust hybrid evolutionary method for ARMA model identification

A new hybrid evolutionary method is proposed. This method alleviates the dependency of pure evolutionary algorithms on the complexity of a given time series and turns out to be very reliable in identifying the correct order and estimation the true parameters' values of a given system model. It combines the effectiveness of the multi-model partitioning theory with the robustness of evolutionary algorithms. Although the system structure is a bit complicated, simulation results show that the proposed method gives better results compared to the conventional multi-model adaptive filter algorithm and the pure evolutionary ones, since it has not only the ability to perform well in searching the whole parameter space, but also to cope with the complexity of the model and reliably lead to the correct order and the true parameters' values. The method can be implemented in a parallel environment thus increasing the computational speed.