Pattern Matching and Neural Networks based Hybrid Forecasting System
Fieldsend, Jonathan E.
Lecture Notes in Computer Science
Springer Berlin Heidelberg
In this paper we propose a Neural Net-PMRS hybrid for forecasting time-series data. The neural network model uses the traditional MLP architecture and backpropagation method of training. Rather than using the last n lags for prediction, the input to the network is determined by the output of the PMRS (Pattern Modelling and Recognition System). PMRS matches current patterns in the time-series with historic data and generates input for the neural network that consists of both current and historic information. The results of the hybrid model are compared with those of neural networks and PMRS on their own. In general, there is no outright winner on all performance measures, however, the hybrid model is a better choice for certain types of data, or on certain error measures.
Copyright © 2001 Springer-Verlag Berlin Heidelberg. The final publication is available at link.springer.com
Book title: Advances in Pattern Recognition — ICAPR 2001
Second International Conference on Advances in Pattern Recognition (ICAPR 2001), Rio de Janeiro, Brazil, March 11–14, 2001
Volume 2013, pp. 72-82