| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To efficiently detect abnormality of traffic, such as the amount of data transmitted and received through a network, with high precision. <P>SOLUTION: The embodiment of the present invention is to predict a reference value for deciding abnormality of the amount of data when the amount of data totaled in time series in predetermined time interval units periodically increases and decreases. A selector 3 divides data of one cycle in time series in preset time slot units and outputs the divided data, parameter estimating units 4a to 4n estimate parameters for predicting a reference value for data in respective time slots output from the selector 3 by using a first time-series analysis method such as an EM algorithm, and calculate weighted average values of the estimated parameter estimated values and parameter estimated values in identical time slots of a past cycle as parameters of a next cycle, and predictors 6a to 6n calculate reference values by a second time-series analysis method using a time-series analytic model such as a Kalman filter using the parameters in the next cycle calculated by the respective parameter estimating units 4a to 4n. <P>COPYRIGHT: (C)2009,JPO&INPIT</p> |
