TIME SERIES DATA ANALYSIS METHOD, TIME-SERIES DATA ANALYZER AND COMPUTER PROGRAM

• Main Authors: MAYA SHIGERU, INAGI TATSUYA, YAMAGUCHI AKIHIRO, MARUCHI KOHEI
• Format: Patent
• Language: English; Japanese
• Published: 15.10.2020
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

To learn model parameters and feature waveforms of a classification model that is effective for classifying time-series data with high accuracy.SOLUTION: The time-series data analysis method as an embodiment of the present invention includes a feature vector generation step for generating a plurality of first feature vectors including feature quantities of the plurality of feature waveforms based on a plurality of first time-series data, which is time-series data belonging to the first class and a distances from the plurality of feature waveforms, and generating a plurality of second feature vectors including feature quantities of the plurality of feature waveforms based on a plurality of second time-series data, which is time-series data belonging to the second class and a distances from the plurality of feature waveforms, and an update processing step for updating a model parameter including weights of the plurality of feature waveforms, and the plurality of feature waveforms based on the plurality of first feature vectors, the plurality of second feature vectors, and a performance index parameter which is a parameter related to the performance index of the classification model.SELECTED DRAWING: Figure 1 【課題】時系列データを高精度にクラス分類するのに有効なクラス分類モデルのモデルパラメータ及び特徴波形を学習する。【解決手段】本発明の実施形態としての時系列データ分析方法は、第１クラスに属する時系列データである複数の第１時系列データと、複数の特徴波形との距離に基づき、前記複数の特徴波形の特徴量を含む複数の第１特徴ベクトルを生成し、第２クラスに属する時系列データである複数の第２時系列データと、前記複数の特徴波形との距離に基づき、前記複数の特徴波形の特徴量を含む複数の第２特徴ベクトルを生成する、特徴ベクトル生成ステップと、前記複数の第１特徴ベクトルと、前記複数の第２特徴ベクトルと、クラス分類モデルの性能指標に関するパラメータである性能指標パラメータとに基づき、前記複数の特徴波形の重みを含むモデルパラメータと、前記複数の特徴波形とを更新する更新処理ステップとを備える。【選択図】図１