摘要 |
PURPOSE:To efficiently extract and hold only an efficient pattern when the pattern is extracted from past time-series data. CONSTITUTION:A pattern extraction part 4 groups user-specified data for learning, which are taken out of a data base 2 for learning by a learning data extraction part 3, by a clustering method, discards clusters, which do not reach the necessary number of elements as pattern extraction conditions set by a pattern extraction condition setting part 7 among the respective clusters, as exceptions, and stores only clusters which meet the conditions of the necessary number of elements as patterns in a pattern data base 5. |