Multi-stage algorithm based on neural network committee for prediction and search for precursors in multi-dimensional time series

Abstract

The studied problem is prediction of time series based on preceding values of several time series (a multi-dimensional time series). Besides prediction itself, the task is finding precursors, i.e. determination of a set of the most significant input features in coordinates "initial time series - lag". A four-stage prediction algorithm based on neural network committee has been suggested, implemented and studied. The algorithm has been successfully tested on one model problem and on one real world problem. © 2009 Springer Berlin Heidelberg.