| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To obtain a neural network system for time series data prediction that can calculate an accurate predictive value, while interrelating analysis levels. <P>SOLUTION: A neural network system 100 for time series data prediction which processes time series data representing numerical values of a prediction target in time series to calculate predictive values of the numerical values, includes an input unit 110, which receives values showing features in a plurality of analysis levels which are obtained by multiple resolution analysis of the time series data as analyzes data; and a processing unit 120, including an input layer processing part 121 which, when calculating predictive values, performs processing, based on analyzed data in the highest analysis level from among a plurality of analysis levels and processes data resulting from the processing, together with analyzed data in a next highest level and performs processing, based on both of data resulting from processing based on analyzed data at one analysis level and analyzed data at an analysis level lower than the one analysis level, to perform processing from the highest analysis level up to the lowest analysis level. <P>COPYRIGHT: (C)2011,JPO&INPIT</p> |
