| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To provide an electricity price zone prediction method capable of highly precise prediction. <P>SOLUTION: At first, initialization setting of each prediction model 1 to i is carried out (step S1). Then, an electricity price zone is divided into n pieces of zones, and n-1 pieces of thresholds as the boundaries of the respective zones are determined (step S2). Also, learning of the prediction models 1 to i is performed by using a steepest descent method (step S3). Then, input variables are applied to the respective prediction models 1 to i learnt by the step S3 as input data this time, and the binary prediction of whether or not the input variables are more than the thresholds set by the step S1 is performed (step S4). Finally, a price range is determined by using a binary tree by using the prediction result of the step S4. <P>COPYRIGHT: (C)2008,JPO&INPIT</p> |
