Munich Personal RePEc Archive

Forecasting international bandwidth capacity using linear and ANN methods

Madden, Gary G and Tan, Joachim (2008): Forecasting international bandwidth capacity using linear and ANN methods. Published in: Applied Economics No. 40 (2008): pp. 1775-1787.

[img]
Preview
PDF
MPRA_paper_13005.pdf

Download (145Kb) | Preview

Abstract

An artificial neural network (ANN) can improve forecasts through pattern recognition of historical data. This article evaluates the reliability of ANN methods, as opposed to simple extrapolation techniques, to forecast Internet bandwidth index data that is bursty in nature. A simple feedforward ANN model is selected as a nonlinear alternative, as it is flexible enough to model complex linear or nonlinear relationships without any prior assumptions about the data generating process. These data are virtually white noise and provides a challenge to forecasters. Using standard forecast error statistics, the ANN and the simple exponential smoothing model provide modestly better forecasts than other extrapolation methods

UB_LMU-Logo
MPRA is a RePEc service hosted by
the Munich University Library in Germany.