AUTOREGRESSIVE MODEL LEARNING DEVICE FOR TIME SERIES DATA AND DEVICE FOR DETECTING DEVIATED VALUE AND CHANGING POINT USING THE SAME

摘要

PPROBLEM TO BE SOLVED: To precisely detect a deviated value and a changing point by measuring the degree showing the statistically deviated value and the changing point appearing in a time series of data. PSOLUTION: As to a series of data inputted sequentially, a time series model learning device 21 learning a generating mechanism of a data sequence read as a time series statistical model, a score calculation device 23 calculating a deviated value score of each data based on a parameter of a time series model and an inputted data, a moving-average calculation device 22 calculating the moving average of the deviated value score, a time series model learning device 24 learning a generating mechanism of a sequence of the moving average as a time series statistical model, and the score calculation device 23 further calculating the deviated value score of the moving average based on the moving average of the deviated value score and outputting it as a changing point score of the original data are combined for calculating the deviated value score and the changing point score to detect the deviated value and the changing point. PCOPYRIGHT: (C)2004,JPO