Multi-step Ahead Forecasting Using Cartesian Genetic Programming

This paper describes a forecasting method that is suitable for long range predictions. Forecasts are made by a calculating machine of which inputs are the actual data and the outputs are the forecasted values. The Cartesian Genetic Programming (CGP) algorithm finds the best performing machine out of a huge abundance of candidates via evolutionary strategy. The algorithm can cope with non-stationary highly multivariate data series, and can reveal hidden relationships among the input variables. Multiple experiments were devised by looking at several time series from different industries. Forecast results were analysed and compared using average Symmetric Mean Absolute Percentage Error (SMAPE) across all datasets. Overall, CGP achieved comparable to Support Vector Machine algorithm and performed better than Neural Networks.

[1]  Yen-Ming Chiang,et al.  Multi-step-ahead neural networks for flood forecasting , 2007 .

[2]  Julian Francis Miller,et al.  Principles in the Evolutionary Design of Digital Circuits—Part II , 2000, Genetic Programming and Evolvable Machines.

[3]  Paulo Cortez,et al.  Sensitivity analysis for time lag selection to forecast seasonal time series using Neural Networks and Support Vector Machines , 2010, The 2010 International Joint Conference on Neural Networks (IJCNN).

[4]  Yang Liu,et al.  Automatic Code Generation on a MOVE Processor Using Cartesian Genetic Programming , 2010, ICES.

[5]  J. Scott Armstrong,et al.  Long-Range Forecasting: From Crystal Ball to Computer , 1981 .

[6]  Julian Francis Miller.  Cartesian Genetic Programming , 2011, Cartesian Genetic Programming.

[7]  Zhongyi Hu,et al.  Beyond One-Step-Ahead Forecasting: Evaluation of Alternative Multi-Step-Ahead Forecasting Models for Crude Oil Prices , 2013, ArXiv.

[8]  Lukás Sekanina,et al.  Gate-level optimization of polymorphic circuits using Cartesian Genetic Programming , 2009, 2009 IEEE Congress on Evolutionary Computation.

[9]  Douglas M. Hawkins,et al.  The Problem of Overfitting , 2004, J. Chem. Inf. Model..

[10]  Amaury Lendasse,et al.  Methodology for long-term prediction of time series , 2007, Neurocomputing.

[11]  Ibrahim El-Amin,et al.  Artificial neural networks as applied to long-term demand forecasting , 1999, Artif. Intell. Eng..

[12]  Dobrivoje Popovic,et al.  Computational Intelligence in Time Series Forecasting: Theory and Engineering Applications (Advances in Industrial Control) , 2005 .

[13]  Amir F. Atiya,et al.  A review and comparison of strategies for multi-step ahead time series forecasting based on the NN5 forecasting competition , 2011, Expert Syst. Appl..

[14]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[15]  Farid Atry,et al.  Multi-step ahead forecasts for electricity prices using NARX: A new approach, a critical analysis of one-step ahead forecasts , 2009 .