UDK 004.94
D. V. Fedotov, E. S. Semenkin
Siberian State Aerospace University named after academician M. F. Reshetnev 31, Krasnoyarsky Rabochy Av., Krasnoyarsk, 660014, Russian Federation E-mail: fedotov.dm.v@gmail.com, eugenesemenkin@yandex.ru
Time series prediction is quite complex and interesting problem. Usually a forecast is based on information from the past, i.e. on previous values of features. The economic situation in the country depends on different factors and events and influences many situations. The innovation projects in aerospace field require a large amount of investments and take a long time for development and implementation. These kinds of investments require relevant and reliable information about possible future economic situation to minimize risks and correct the plans in right time. In particular, to get such information one may use the time series of the various indexes of economic development. In this paper, the neural network predictors are used as the main technology of forecasting. They are generated automatically by means of genetic algorithms that allow increasing the quality of the prediction. Besides, the advantages of the use of self-configuring genetic algorithms in comparison with standard genetic algorithms are demonstrated when adjusting the weighting coefficients of the neural network. The quality of the solution provided by the neural network depends also on its structure. In this paper, the genetic programming is used for the neural network structure design due to its ability to provide effective and flexible models without a human expert. This allows automating the neural network based predictors configuring. The combination of these intellectual information technologies is tested on the financial time series forecasting task using the values of the prices and volumes along with the technical indicators as the inputs. The developed system allows automated designing the neural network based predictors and getting a high quality forecast of economic indexes changes. The developed system shows better results in the comparison with the other forecasting technologies.
neural network, evolutionary algorithms, self-configuring, time series, forecasting
  1. Ossovskiy S. Neyronnye seti dlya obrabotki informatsii [Neural networks for information processing]. Moscow, Finansy i statistika Pulb., 2002, 344 p.

  2. Anderson D., McNeill G. Artificial neural networks technology. DACS report. 1992, 87 p.

  3. Haupt R. L., Haupt S. E. Practical Genetic Algorithms, 2ed., Wiley, 2004. P. 261.

  4. Koza J. R. Genetic Programming: On the Programming of Computers by Means of Natural Selection, MIT Press, 1992.

  5. Galushkin A. I. Sintez mnogosloynykh sistem raspoznavaniya obrazov [Synthesis of the multilayer pattern recognition systems]. Moscow, Energiya Pulb., 1974.

  6. Werbos P. J., Beyond regression: New tools for prediction and analysis in the behavioral sciences. Ph. D. thesis, Harvard University, Cambridge, MA, 1974.

  7. Wright A. Genetic algorithms for real parameter optimization. Foundations of Genetic Algorithms. San Mateo, CA: Morgan Kaufmann, 1991, p. 205–218.

  8. Eiben A. E., Smith J. E., Introduction to evolutionary computing. Springer, 2003, Berlin, Germany.

  9. Semenkin E., Semenkina M. Self-configuring genetic algorithm with modified uniform crossover operator, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 7331 LNCS (PART 1), 2012, p. 414–421.

  10. Gomez J. Self Adaptation of Operator Rates Evolutionary Algorithms, Genetic and Evolutionary Computation, GECCO 2004. Lecture Notes in Computer Science. 2004, Vol. 3102, p. 1162–1173.

  11. Semenkina M., Semenkin E. Hybrid Self-configuring Evolutionary Algorithm for Automated Design of Fuzzy Classifier, Advances in Swarm Intelligence. Y. Tan et al. (Eds.): ICSI 2014, Part 1, LNCS 8794, 2014, p. 310–317.

  12. Burakov S. V., Semenkin E. S. Solving variational and Cauchy problems with self-configuring genetic programming algorithms, International Journal of Innovative Computing and Applications. 2013, Vol. 5, Iss. 3, p. 152–162.

  13. Semenkin E., Semenkina M. Classifier Ensembles Integration with Self-Configuring Genetic Programming Algorithm, Adaptive and Natural Computing Algorithms. Lecture Notes in Computer Science, 2013, Vol. 7824. Springer-Verlag, Berlin, Heidelberg, p. 60–69.

  14. Molga M., Smutnicki Cz., 2005. Test functions for optimization need. 3 kwietnia.

  15. Fama E. F. Random walks in stock market prices, Financial Analysts Journal. 1965, no. 21(5), p. 55–59.

  16. Cootner Paul H. The random character of stock market prices. MIT Press. 1964.

  17. R2013b – Updates to the MATLAB and Simulink product families, Available at: http://www.mathworks. com/products/new_products/release2013b.html (accessed 18.01.2015).

  18. Zhang Y. Prediction of Financial Time Series with Hidden Markov Models, Simon Fraser University, 2004.

  19. Fedotov D. V. On financial time series prediction using neural network models and evolutionary algorithms // Materials of the Third International Workshop on Mathematical Models and its Applications (IWMMA-2014), 2014, P. 3.

  20. Predictive Analytics, Data Mining, Self-service, Open source – RapidMiner. Available at: http:// rapidminer.com/ (accessed 20.10.2014).

  21. Rapid-I Market Place Weka Extension for RapidMiner, Available at: http://marketplace.rapid-i.com/UpdateServer/faces/product_details.xhtml?productId=rmx_weka (accessed 20.10.2014).

Fedotov Dmitrii Valer'evich – laboratory research worker, Siberian State Aerospace University named after academician M. F. Reshetnev. E-mail: fedotov.dm.v@gmail.com

Semenkin Evgeny Stanislavovich – Dr. Sc., Professor, Professor of System analysis and operations research department, Siberian State Aerospace University named after academician M. F. Reshetnev. E-mail: eugenesemenkin@yandex.ru