FIELD: artificial neural network.
SUBSTANCE: invention relates to predicting time series values using an artificial neural network (ANN). The invention can be used to solve a wide range of problems where it is required to predict new values of the studied characteristic from its known values in the past. The method and system for forecasting a time series using an artificial neural network includes: a method for creating extended metadata series, a method for encrypting data for constructing optimal training samples for an ANN, a method and a training subsystem for an artificial neural network, a subsystem for ANN operation in a forecast mode, a method for constructing a forecast with using an ANN that has passed the training procedure.
EFFECT: invention improves accuracy of predicting time series values of arbitrary characteristics.
5 cl, 7 dwg
