PREDICTION SYSTEM, MODEL GENERATION SYSTEM, METHOD, AND PROGRAM

Title:

PREDICTION SYSTEM, MODEL GENERATION SYSTEM, METHOD, AND PROGRAM

Document Type and Number:

WIPO Patent Application WO/2019/187372

Kind Code:

A1

Abstract:

A model generation device 60 of the present invention is provided with: a regularization parameter candidate setting means 61 which outputs a regularization parameter search set for which a plurality of solution candidates are set, the solution candidates having a plurality of mutually different values of at least a regularization parameter that affects the term of a strong regularization variable that is one or a plurality of variables specifically determined among explanatory variables; a model learning means which learns a prediction model corresponding to each of the plurality of solution candidates; an accuracy evaluation means 62 which evaluates prediction accuracy of each prediction model that has been learned; a transition evaluation means 63 which, with respect to each prediction model that has been learned, evaluates a defective sample number that is the number of samples in which graph shape or the transition is not valid; and a model determination means 64 which, on the basis of the result of evaluation concerning prediction accuracy and the result of evaluation concerning graph shape or the defective sample number, determines a prediction model to be used for progress prediction.

Inventors:

TANIGUCHI ATSUSHI (JP)
TAMANO HIROSHI (JP)

Application Number:

PCT/JP2018/045610

Publication Date:

October 03, 2019

Filing Date:

December 12, 2018

Export Citation:

Click for automatic bibliography generation Help

Assignee:

NEC SOLUTION INNOVATORS LTD (JP)

International Classes:

G16H50/50

Domestic Patent References:

WO2017098631A1

2017-06-15

Foreign References:

JP2018010475A	2018-01-18
JP2016146046A	2016-08-12
JP2012064087A	2012-03-29
JP2014225175A	2014-12-04
JP2018068278A	2018-05-10

Other References: