STATE TRANSITION PREDICTION MODEL LEARNING APPARATUS, METHOD, AND PROGRAM

To update a state transition prediction model for use in calculation of a score representing magnitude of tendency in occurence of state transition to be inspected, by using terminated training data.SOLUTION: Under the control of a data kind discriminating unit 1121, it is discriminated that acquire...

Full description

Saved in:
Bibliographic Details
Main Authors YOSHIDA KAZUHIRO, CHIBA AKIHIRO, ASANOMA NAOKI, YAMADA TOMOHIRO, HAYASHI KATSUYOSHI, YABUUCHI TSUTOMU, FUJINO AKINORI, KURASAWA HIROSHI, EGUCHI KANA, AZUMA SHOZO
Format Patent
LanguageEnglish
Japanese
Published 22.07.2019
Subjects
CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES
PHYSICS
SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR
Online AccessGet full text

Cover

More Information
Summary:To update a state transition prediction model for use in calculation of a score representing magnitude of tendency in occurence of state transition to be inspected, by using terminated training data.SOLUTION: Under the control of a data kind discriminating unit 1121, it is discriminated that acquired training data is completion training data and terminated training data. Under the control of a model learning unit 113, two records are selected from each of training data. For example, only one of the two records is a record from the terminated training data. If a predetermined condition is satisfied, under the control of a transition prediction unit 114, a score of each of the two records is calculated by using a first prediction model. Under the control of a model learning unit 113, the first prediction model is updated so that magnitude of the calculated scores should correspond to length of a period to occurrence of a state transition of the two records calculated under the control of a transition period calculating unit 1122.SELECTED DRAWING: Figure 1 【課題】検査対象の状態遷移が発生する傾向の大きさを表すスコアを算出するための状態遷移予測モデルを、打ち切り訓練データを利用して更新できるようにする。【解決手段】データ種判別部１１２１の制御の下、取得された訓練データが完了訓練データと打ち切り訓練データのいずれであるかが判別される。モデル学習部１１３の制御の下、各訓練データから２つのレコードが選択され、例えば、当該２つのレコードの一方のみが打ち切り訓練データからのレコードであり、所定の条件が満たされた場合、遷移予測部１１４の制御の下、上記２つのレコードの各々について、第１予測モデルを使用してスコアが算出される。モデル学習部１１３の制御の下、当該算出されたスコアの大小関係が、当該２つのレコードについて遷移期間算出部１１２２の制御下で算出された状態遷移が発生するまでの期間の長さの大小関係に対応するように、第１予測モデルが更新される。【選択図】図１
Bibliography:Application Number: JP20180000147