INFORMATION PROCESSING SYSTEM AND INFORMATION PROCESSING METHOD

To shorten the calculation time required to predict the quantity related to a prediction object.SOLUTION: An information processing system 100 stores history information in which calculation time required for predicting the quantity related to a prediction object and adoption/non-adoption combinatio...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: ARA HIROMI, OJIRO DAICHI, WATANABE TAKASHI, HIRAYAMA JUNICHI
Format: Patent
Sprache:eng ; jpn
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:To shorten the calculation time required to predict the quantity related to a prediction object.SOLUTION: An information processing system 100 stores history information in which calculation time required for predicting the quantity related to a prediction object and adoption/non-adoption combinations of a plurality of time series prediction methods used for predicting the quantity related to the prediction object are associated with each other, generates a new adoption/non-adoption combination that is not present in the history information, generates a calculation time estimation model with the calculation time as an objective variable and an existing adoption/non-adoption combination stored in the history information as an explanatory variable, outputs estimated calculation time for each adoption/non-adoption combination by inputting the existing adoption/non-adoption combination and a new combination into the calculation time estimation model, and searches for a first adoption/non-adoption combination that falls within an allocation calculation time assigned to the prediction object among the estimated calculation times, as a specific adoption/non-adoption combination, from the adoption/non-adoption combinations.SELECTED DRAWING: Figure 1 【課題】予測対象に関する数量の予測に要する計算時間の短縮化を図ること。【解決手段】情報処理システム100は、予測対象に関する数量の予測に要した計算時間と、予測対象に関する数量の予測に用いられた複数の時系列予測手法の採否の組み合わせと、を関連付けた履歴情報を記憶しており、履歴情報には存在しない新規な採否の組み合わせを生成し、計算時間を目的変数とし、履歴情報に記憶されている既存の採否の組み合わせを説明変数とした計算時間推定モデルを生成し、計算時間推定モデルに既存の採否の組み合わせおよび新規な組み合わせをそれぞれ入力することにより、採否の組み合わせごとの推定計算時間を出力し、採否の組み合わせの中から、推定計算時間のうち予測対象に割り当てられた割当計算時間に収まる第1の採否の組み合わせを特定の採否の組み合わせとして探索する。【選択図】図1