A Machine Learning Approach to Define Weights for Linear Combination of Forecasts

The linear combination of forecasts is a procedure that has improved the forecasting accuracy for different time series. In this procedure, each method being combined is associated to a numerical weight that indicates the contribution of the method in the combined forecast. We present the use of machine learning techniques to define the weights for the linear combination of forecasts. In this paper, a machine learning technique uses features of the series at hand to define the adequate weights for a pre-defined number of forecasting methods. In order to evaluate this solution, we implemented a prototype that uses a MLP network to combine two widespread methods. The experiments performed revealed significantly accurate forecasts.

[1]  Teresa Bernarda Ludermir,et al.  Using Machine Learning Techniques to Combine Forecasting Methods , 2004, Australian Conference on Artificial Intelligence.

[2]  Teresa Bernarda Ludermir,et al.  Meta-learning approaches to selecting time series models , 2004, Neurocomputing.

[3]  James L. Powell,et al.  Time Series Models , 2021, Stochastic Limit Theory.

[4]  Sevket I. Gunter,et al.  An empirical analysis of the accuracy of SA, OLS, ERLS and NRLS combination forecasts , 1992 .

[5]  C. Granger,et al.  Improved methods of combining forecasts , 1984 .

[6]  Howard B. Demuth,et al.  Neutral network toolbox for use with Matlab , 1995 .

[7]  Ricardo Vilalta,et al.  Introduction to the Special Issue on Meta-Learning , 2004, Machine Learning.

[8]  Jeffrey Sohl,et al.  An intelligent model selection and forecasting system , 1999 .

[9]  Francisco de A. T. de Carvalho,et al.  A Modal Symbolic Classifier for selecting time series models , 2004, Pattern Recognit. Lett..

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

[11]  Lutz Prechelt,et al.  A Set of Neural Network Benchmark Problems and Benchmarking Rules , 1994 .

[12]  Fred Collopy,et al.  Automatic Identification of Time Series Features for Rule-Based Forecasting , 2001 .

[13]  Bay Arinze,et al.  Selecting appropriate forecasting models using rule induction , 1994 .

[14]  J. Armstrong,et al.  Findings from Evidence-Based Forecasting: Methods for Reducing Forecast Error , 2006 .

[15]  Benito E. Flores,et al.  Use of the sign test to supplement the percentage better statistic , 1986 .

[16]  Geoffrey E. Hinton,et al.  Learning representations by back-propagating errors , 1986, Nature.

[17]  T. Evgeniou,et al.  To combine or not to combine: selecting among forecasts and their combinations , 2005 .

[18]  Chao-Hsien Chu,et al.  Neural network system for forecasting method selection , 1994, Decis. Support Syst..

[19]  Derek W. Bunn,et al.  Review of guidelines for the use of combined forecasts , 2000, Eur. J. Oper. Res..