On the basis of genetic algorithm, an intelligent search approach to determination of parameters of ARMAX(Autor Regressive Moving Average model with external input) processes was proposed. By representing the system with pole and zero pairs and repairing illegal chromosomes, the search space is limited to stable schemes. In calculation of objective function the “shifted data window” was designed, so that every input output pair is used to guide the evolution and the “Data Saturation” is avoided. To prevent premature convergence, the adaptive fitness function was introduced, the conventional crossover and mutation operator was modified and the “catastrophic mutation” which is based on Metropolis mechanism was adopted. So the performance of convergence to the global optimum is improved. The validity and efficiency of proposed algorithm were illustrated by simulated results. On the basis of genetic algorithm, an intelligent search approach to determine of parameters of ARMAX (Autor Regressive Moving Average model with external input) processes was proposed. By presenting the system with pole and zero pairs and repairing illegal chromosomes, the search space is limited to stable schemes. “In calculation of objective function the ” shifted data window “was designed, so that every input output pair is used to guide the evolution and the” Data Saturation “is avoided. To prevent premature convergence, the adaptive fitness function was introduced, the conventional crossover and mutation operator was modified and the ”catastrophic mutation " which is based on Metropolis mechanism was adopted. So the performance of convergence to the global optimum is improved. The validity and efficiency of proposed algorithm were were illustrated by simulated results.