PURPOSE: To prevent disable calculation even in the presence of redundancy in calender information and to apply the method even to a time series having a trend or a stepwise time series fluctuation by reducing a storage capacity considerably and decreasing the calculation processing time.
CONSTITUTION: A state vector of n-th day comprising a weight coefficient for each attribute information represented calender information of n-th day used for estimating a communication traffic of n-th day and comprising a component of time series fluctuation representing time series fluctuation of a daily communication traffic before the n-th day is revised based on a prediction error between a statistic quantity of the communication traffic measured by the n-th day and an n-th day prediction traffic communication quantity estimated before the measurement of the n-th day communication traffic and the statistic quantity of the communication traffic of (n+1)th day is estimated based on the revised n-th day state vector.