Using Ensembles Of Neural Networks With Different Scales Of Input Data For The Analysis Of Telemetry Data

This article gives a brief description of the main methods of forming parallel ensembles of experts, in particular ensembles of neural networks. Also the learning algorithm of neural network ensembles with elements of the evolution strategy described. The problem of concept drift and methods of its solution using incremental learning ensembles of experts described. Also the method of searching the important features of the time series for forecasting is given. And the model of two-level learning algorithm of neural network ensembles with different time scale is proposed for the prediction of telemetry data. This model is based on an ensemble of neural networks that use mentioned algorithm of learning. Also presents the results of tests of the model on the task of forecasting multivariate time series of telemetry.