The accuracy of a procedural approach to specifying feedforward neural networks for forecasting

The comparative accuracy of feedforward neural networks (NN) when applied to time series forecasting problems remains uncertain. This is because most studies suffer from either of two defects--they choose the NN from a wide range of alternatives in order to present the forecast accuracy results in the best light or they do not compare the results with suitable benchmarks. In order to overcome both these objections, this paper proposes an objective procedure for specifying a feedforward NN model and evaluates its effectiveness by examining its forecasting performance compared with established benchmarks. After the selection of input nodes based on cross-validation, a three-stage procedure is proposed here which consists of sequentially selecting first the learning rate followed by the number of nodes and the initial weights. This paper shows that NNs only perform robustly if they are built by considering these three factors jointly. In an empirical demonstration of the strength of the approach, those NN models, built by considering all three factors, performed better than other competitive statistical methods when evaluated rigorously on a standard test data set.

[1]  Michael Y. Hu,et al.  A simulation study of artificial neural networks for nonlinear time-series forecasting , 2001, Comput. Oper. Res..

[2]  Murray Smith,et al.  Neural Networks for Statistical Modeling , 1993 .

[3]  Emanuel Parzen,et al.  ARARMA models for time series analysis and forecasting , 1982 .

[4]  J. Scott Armstrong,et al.  Research Report - Principles for Examining Predictive Validity: The Case of Information Systems Spending Forecasts , 1994, Inf. Syst. Res..

[5]  Robert Fildes,et al.  Generalising about univariate forecasting methods: Further empirical evidence , 1998 .

[6]  Fred L. Collopy,et al.  Error Measures for Generalizing About Forecasting Methods: Empirical Comparisons , 1992 .

[7]  Robert Fildes,et al.  The evaluation of extrapolative forecasting methods , 1992 .

[8]  Geoffrey E. Hinton,et al.  Learning internal representations by error propagation , 1986 .

[9]  Robert L. Winkler,et al.  The Accuracy of Extrapolation (Time Series) Methods , 1982 .

[10]  C. Chatfield,et al.  Neural networks: Forecasting breakthrough or passing fad? , 1993 .

[11]  Chris Chatfield,et al.  Time series forecasting with neural networks , 1998, Neural Networks for Signal Processing VIII. Proceedings of the 1998 IEEE Signal Processing Society Workshop (Cat. No.98TH8378).

[12]  Michael Y. Hu,et al.  Forecasting with artificial neural networks: The state of the art , 1997 .

[13]  Werner A. Stahel,et al.  Forecasting demand for special telephone services: A case study , 1990 .

[14]  Michael Y. Hu,et al.  A Cross‐Validation Analysis of Neural Network Out‐of‐Sample Performance in Exchange Rate Forecasting , 1999 .

[15]  Spyros Makridakis,et al.  The M3-Competition: results, conclusions and implications , 2000 .

[16]  E. S. Gardner,et al.  Forecasting Trends in Time Series , 1985 .

[17]  William Remus,et al.  Neural Network Models for Time Series Forecasts , 1996 .

[18]  J. van Leeuwen,et al.  Neural Networks: Tricks of the Trade , 2002, Lecture Notes in Computer Science.

[19]  Philipp Slusallek,et al.  Introduction to real-time ray tracing , 2005, SIGGRAPH Courses.

[20]  Ralph Neuneier,et al.  How to Train Neural Networks , 1996, Neural Networks: Tricks of the Trade.

[21]  Heekuck Oh,et al.  Neural Networks for Pattern Recognition , 1993, Adv. Comput..

[22]  Fred Collopy,et al.  How effective are neural networks at forecasting and prediction? A review and evaluation , 1998 .

[23]  Robert L. Winkler,et al.  The accuracy of extrapolation (time series) methods: Results of a forecasting competition , 1982 .

[24]  J. Keith Ord,et al.  Automatic neural network modeling for univariate time series , 2000 .

[25]  Spyros Makridakis,et al.  The M3-Competition1 , 2000 .