Using Ensembles of Neural Networks for Forecasting Telemetry Data

Abstract

In this paper we propose an approach to solving the problems of forecasting multivariate time series telemetry data that describe the state of small airborne objects. The main objective of the proposed method - it's automated design and development of neural network models for solving such problems, namely the choice of model parameters are close to optimal. The approach is based on the use of ensembles of neural networks. In this case learning algorithm uses some elements of evolutionary strategy. The article also describes the experiments and experimental data.