摘要 |
Many actual data 30, each consisting of a set of an actual water distribution amount obtained in advance from a water management system 10 and a variable needed for prediction, are loaded and retained as history data 52, an input space of a case base 54 is quantized according to a desired output allowance error, each history data 52 is set in the quantized input space, and a case representing at least one history data is prepared to generate a case base 52. Then a similar case corresponding to a newly input prediction condition 40 is retrieved from the case base 54, and a water distribution amount corresponding to the prediction condition 40 is estimated from the retrieved similar case.
|