| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To make high-speed and high-precision predictions based on a small number of cases by optimizing a set of known cases accumulated in the past, without deteriorating predicting accuracy. <P>SOLUTION: A prediction support apparatus 200 creates a set (x) of cases for each known case included in a number of cases in a set of initial known cases, in which set (x) known cases (hereinafter referred to as 'deleted cases') are deleted from a set of known cases given importance and the importance given to the deleted cases is added to the most similar case of the deleted cases. Then the minimum value of the total sum (C) of differences between a similarity sum (A) for each desired variable value such that the set of initial known cases is calculated as a set of known examples and a set of unknown cases and a similarity sum (B) such that a set of initial known cases is calculated as a set of unknown cases while the set (x) of cases is regarded as a set of known cases. The deleted case corresponding to the total sum (C) is determined as the subject of deleting and deleted from the set of known cases given importance. <P>COPYRIGHT: (C)2010,JPO&INPIT</p> |
