| 摘要 |
PROBLEM TO BE SOLVED: To provide a data set selecting device for applying an experiment design by an active learning method to a plurality of items of data to which data set is defined in advance. SOLUTION: The selecting device comprises a data input reception unit 102 receiving a data set of training data including a plurality of attribute values and corresponding known output values and a data set of predictive data to obtain unknown output values corresponding to a plurality of prepared attribute values; a prediction result reception unit 110 learning a correspondence relationships of the output values corresponding to a plurality of attribute values of training data by using an active learning device 30 including a plurality of prediction algorithms, predicting the output values corresponding to the predictive data by using a plurality of learned correspondence relationships, and obtaining a plurality of prediction result values of the plurality of prediction algorithms; and an extraction unit 116 selecting a data set with large variations of the plurality of prediction results by the plurality of obtained prediction algorithms among the data sets of the corresponding predictive data and extracting the selected predictive data set as experiment candidate data. COPYRIGHT: (C)2008,JPO&INPIT
|
