摘要 |
<p><P>PROBLEM TO BE SOLVED: To provide a traffic volume prediction apparatus, a program, and a method capable of predicting an upper limit value of traffic volume variation. <P>SOLUTION: The traffic volume prediction apparatus includes an actual traffic database, a weighting factor memory, a multiple regression processing portion, and a weighting factor determination portion. A past traffic volume is associated with its time attribute information, and they are accumulated in the database. A weighting factor w<SB>i</SB>in an objective function is accumulated in the weighting factor memory. The multiple regression processing portion derives a regression equation by multiple regression analysis, in which minimization of a squared error with a weight is made as an objective function, and outputs a learned traffic volume. The weighting factor determination portion increments the weighting factor w<SB>i</SB>according to a difference between an actual traffic volume y<SB>i</SB>and a predicted traffic volume z<SB>i</SB>, only if y<SB>i</SB>is larger than z<SB>i</SB>. As for each of actual traffic volumes y<SB>i</SB>, processing of multiple regression processing portion and weighting factor determination portion are controlled to repeat for the number of times of a predetermined, and the regression equation is derived as a result. <P>COPYRIGHT: (C)2008,JPO&INPIT</p> |