The averaged deviations of computed result of the model in the learning process and the deviation of computed results in the controlling process are reacted changeably with rule number and study rate.