| 摘要 |
<p>PURPOSE:To shorten the learning time and facilitate the improvement of the prediction precision by retrieving result data which approximate an observed value by utilizing tree structure and predicting a demand value corresponding to the observed value according to the result. CONSTITUTION:When the observed value is inputted to a demand prediction part 13, a data retrieval part 12 retrieves the data base in a data control part 11 to extract such data that the observed values are the same group of weather, the temperature differences from maximum temperature are <=5 deg., and holiday flags are the same. A data control part 11 controls the result data by using the tree structure. Partial data bases are standardized so that the variance of the maximum temperature and humidity is '1' and the mean is '0', and distances are defined on the partial data bases. Further, the demand prediction part 13 predicts the demand value of the amount of distributed water, etc., corresponding to the given observed value by using the defined partial data bases, distances, and prediction peripheries and outputs the predicted value.</p> |
