PURPOSE: To shorten the time before starting to use a neural net by starting the use of the neural net on condition that the difference between the resulted data obtained through the state of actual usage and the output of the neural net is less than a specified value and that this state continues for specific times.

CONSTITUTION: An inference unit 21, which has learning function of a neural net applied to the assignment control, causes an activity 'a' of respective partial system models to be outputted by inputting a matrix W indicating traffic acceptance and load. A partial model unit 22 causes each model to output a group administrative control response result 'fm' based on the combination of control parameters within a specified range. A group administrative control response inference result 'y' is then found by composing each data 'a' and 'fm' by a composition unit 23 and an inference result evaluation unit 24 forms traffic demand and combination of control parameters. A system regulaton unit 25 gives the inference unit 21 and the partial model unit 22 correct output based on data 'y' and the response result from the inference result evaluation unit 24.