Computational and Applied Math Proseminar

Department of Mathematics, Arizona State University

Tuesday, November 30, 2004, 12:15 p.m. in GWC Room 409

Brad Morantz

Department of Mathematics and Statistics

Time Series Forecasting with a Weighted Window Artificial Neural Network Forecaster

Abstract The presentation begins with a brief introduction into time series forecasting, what they are, why they are used, when to use them, and how to do it. Some standard methodology is described and then weighted windows (WW) is introduced. WW is a contribution to neural networks, as it allows the system to learn in a more realistic and intelligent manner, encompassing the concept of recency learning. Residual has been attenuated by as much as 50% when compared to regular neural networks and other techniques such as random walk and ARIMA.

Over 15 real economic data sets have been forecasted using this method and the results analyzed by both parametric and non parametric statistics.

For further information please contact: mittelmann@asu.edu