Anomaly prediction processing device, anomaly prediction processing method, and anomaly prediction processing program

Abstract

To make it possible to accurately predict a time at which an abnormality occurs with learning processing using a part of time-series data.SOLUTION: When executing learning prediction processing using time series data, a learning prediction device 1 calculates an abnormality degree of each piece of time series data in post-stage prediction processing after generating a classification model parameter in pre-stage learning processing, calculates an approximate equation and a feature equation of the abnormality degree, and then calculates an abnormality occurrence predicted time point and outputs it. Consequently, the learning prediction device 1 is capable of generating the appropriate classification model parameter on the basis of a part of time series data actually obtained from a measurement object, and capable of using an approximate equation and a feature equation of the abnormality degree obtained on the basis of the classification model parameter and other time series data. As a result, the learning prediction device can calculate a quite accurate abnormality occurrence prediction time point.SELECTED DRAWING: Figure 9

Images