METHOD AND DEVICE FOR CLASSIFYING SPATIOTEMPORAL DATA FEATURE AMOUNT

摘要

PROBLEM TO BE SOLVED: To provide a method and device for classifying a spatiotemporal data feature amount with which it is possible to obtain a basis that is less redundant than a conventional objective function, and to raise a data classification rate.;SOLUTION: A method and device for classifying a spatiotemporal data feature amount according to the present invention calculates nonnegative matrix factorization of an observation data matrix Y, generated from time-series data accumulated in data accumulation means, into a basis matrix H and a coefficient matrix U; classifies, on the basis of the decomposed matrix, the time-series data that is the source of the observation data matrix Y into a specific number of clusters; and employs, in the decomposition calculation, a maximum likelihood method that uses a robust function when minimizing the objective function pertaining to an error matrix Y-HU by using a least-square method after attaching a plurality of constraints, the plurality of constraints including those in which a logarithmic function is employed for each matrix element of the basis matrix H and the coefficient matrix U.;COPYRIGHT: (C)2015,JPO&INPIT