Analysis of Multivariate Time Series Model and Neural Networks


Analysis of Multivariate Time Series Model and Neural Networks.


In this work, vector autoregression and neural network approach to multivariate time series analysis is presented. A multilayer perceptron network with backpropagation, gradient descent algorithm has been designed to model the monthly average exchange rates of three major international currencies with respect to naira.

The series span over the period of Jan. 2010 to Aug. 2015. In training the network to learn the combined series of the exchange rates, a remarkable achievement was made. Adding to the beauty of the network model is the fact that the number of units of the input layer was predetermined through the VAR model.

Using some model performance measures (RMSE, MBE and R2), it was recorded that the neural network approach performs better than the VAR model as it yielded minimum error of prediction. However, VAR model is recommended for long term prediction of the exchange rates being that it has a smaller mean bias error.



The need for appropriate decision making against the future occurrences of a system has always been the prime reason behind time series modeling. As decision makers look for appropriate decision tool (model), they invariably try to come up with a predictor which has a minimum associated cost function.

In an effort to get the best model, the analyst always tries to explore every relevant information space so as to get a forecast about the future event which will have relatively minimum error (lutkepohl, 2006). In the study of macroeconomic data, the level of interdependence among contemporaneous variables is always so significant that, it becomes very advantageous to carry the related variables all along so as to extract the structural relationship inherent in them.

The multivariate time series analysis also known vector time series analysis propounded by Sims (1980), has the ability of extracting the dynamic relationship obtainable in multiple equal-spaced-time interrelated variables. Vector autoregressive moving average (VARMA) models are important statistical tools that study the behavior of multiple time dependent variables and forecast future values depending on the history of variations in the data, some other related variables and shocks or innovations.


Ana, B.G. (2012). Empirical macroeconomics-lecture 2. Queen Mary university of London.

Daniya, T. (2015). Forecasting exchange rates using time series analysis: the sample of the currency of Kazakhstan. Cornell university press.

David, E. R., Geoffrey, E. H. and Ronal, J. W. (1986). Learning Representations by Back-propagating errors. Nature Publishing Group.

Eric, S. (2002). Forcasting foreign exchange rates with neural networks. Computer science institute, University of Neuchatel.

Fabio, C. (2007). Methods for Applied Macroeconomic Research. Princeton University press.

Frank, R.J., Davey, N. and Hunt, S.P. (2001). Time series prediction and neural networks. Department of Computer Science, University of Hertfordshire, Hertfield, Uk.

CSN Team.


Comments are closed.

Hey Hi

Don't miss this opportunity

Enter Your Details