Hybrid mathematical models and methods of time series forecasting taking into account external factors

Authors

  • Виталий Николаевич Щелкалин Kharkіv National University of Radioelectronics 14, Lenina str., Kharkiv, Ukraine, 61166, Ukraine

DOI:

https://doi.org/10.15587/1729-4061.2014.31729

Keywords:

forecasting, structural identification, decomposition model, Box-Jenkins method, "Caterpillar"-SSA method

Abstract

The paper presents mathematical models of non-stationary time series forecasting, taking into account external factors and structural identification methods based on the joint use of a multidimensional variant of the "Caterpillar"-SSA method and SARIMAX models, expanded for time series forecasting with several seasonal components and taking into account several exogenous variables.

Structural and parametric identification of the transfer function, connecting the forecast and exogenous time series is performed in the proposed hybrid mathematical models, using formulas for L- or K-continuation of a multi-dimensional variant of the "Caterpillar"-SSA method. Decomposition approach to time series forecasting based on multi-dimensional variant of the "Caterpillar"-SSA method and SARIMAX models lies in decomposition of initial forecast and exogenous time series by the "Caterpillar"-SSA method into multiple time series with simpler structures; forecasting decomposition component data from SARIMAX models and calculating the total forecast, combining forecasts of the constructed simplified models.

The proposed models were tested on electricity and natural gas consumption time series, and the results of their forecasting were compared with the results, obtained by probabilistic SARIMAX models. Experimental results show the high efficiency of the proposed forecasting models in selecting suitable structural parameters in comparison with SARIMAX models.

The results lead to the conclusion that for effective forecasting, it is necessary to perform the decomposition of the studied time series and combine various models, describing both statistical and deterministic components of the time series, which provides better forecasting quality.

Author Biography

Виталий Николаевич Щелкалин, Kharkіv National University of Radioelectronics 14, Lenina str., Kharkiv, Ukraine, 61166

Assistant of Department of Applied Mathematics

References

  1. Evdokimov, A. G., Tevyashev, A. D. (1980). Operativnoe upravlenie potokoraspredeleniem v inzhenernykh setyakh. Khar'kov: Vishcha shkola, 144.
  2. Golyandina, N. E. (2004). Metod «Gusenitsa»-SSA: analiz vremennykh ryadov. Sankt-Peterburg: S.-Peterburgskiy gosudarstvennyy universitet, 74.
  3. Golyandina, N., Nekrutkin, V., Zhigljavsky, A. (2005). Varianty metoda «Gusenitsa»-SSA dlya prognoza mnogomernykh vremennykh ryadov. Trudy IV mezhdunar. konf. Moskva «Identifikatsiya sistem i zadachi upravleniya». Moskow, 1831–1848.
  4. Shchelkalin, V. N. (2014). “Caterpillar”-SSA and Box-Jenkins hybrid models and methods for time series forecasting. Eastern-European Journal of enterprise Technologies, 5/4(71), 43–62. http://journals.uran.ua/eejet/article/view/28172/25654 doi: 10.15587/1729-4061.2014.28172
  5. Lawrance, A. J., Kottegoda, N. T. (1977). Stochasting modelling of riverflow time series. Journal of the Royal Statistical Society. Series A (General), 140 (1), 1–47. doi: 10.2307/2344516
  6. Fernando, D. A. K., Jayawardena, W. A. (1994). Generation and forecasting of monsoon rainfall data. In Proc. of the 20th WEDC conference. Colombo, Sri Lanka, 310–313.
  7. Yurekli, K., Kurunca, A., Ozturkb, F. (2005). Application of linear stochastic models to monthly flow data of Kelkit Stream. Ecological Modelling, 183 (1), 67–75. doi: 10.1016/j.ecolmodel.2004.08.001
  8. Fraedrich, K. (1986). Estimating the dimension of weather and climate attractor. Journal of the Atmospheric Sciences, 43, 419–432.
  9. Vautard, R., Ghil, M. (1989). Singular spectrum analysis in nonlinear dynamics, with applications to paleoclimatic time series. Physica D: Nonlinear Phenomena, 35 (3), 395–424. doi: 10.1016/0167-2789(89)90077-8
  10. Ghil, M., Vautard, R. (1991). Interdecadal oscillations and the warming trend in global temperature time series. Nature, 350 (6316), 324–327. doi: 10.1038/350324a0
  11. Yiou, P., Baert, E., Loutre, M.F. (1996). Spectral analysis of climate data. Surveys in Geophysics, 17 (6), 619–663. doi: 10.1007/bf01931784
  12. Golyandina, N. (2001). Analysis of time series structure: SSA and related techniques. Chapman and Hall/CRC, 320.
  13. Marques, C. A. F., Ferreira, J. A., Rocha, A., Castanheira, J. M., Melo-Goncalves, P., Vaz., N., Dias, J. M. (2005). Singular spectrum analysis and forecasting of hydrological time series. In Meeting of the European-Union-of-Geosciences. Vienna, Austria, 1172–1179.
  14. Hassani, H., Heravi, S., Zhigljavscky, A. (2009). Forecasting European industrial production with singular spectrum analysis. International Journal of Forecasting, 25 (1), 103–118. doi: 10.1016/j.ijforecast.2008.09.007
  15. Shchelkalin, V. N. (2012). Trendovyy podkhod prognozirovaniya vremennykh ryadov na osnove metoda «Gusenitsa»-SSA. Materialy 14-y Mezhdunarodnoy nauchno-tekhnicheskoy konferentsii SAIT. Kiev, 258–259.
  16. Vahabie, A. H., Yousefi, M. M. R., Araabi, B. N., Lucas, C., Barghinia, S. (2007). Combination of Singular Spectrum Analysis and Autoregressive Model for short term load forecasting. IEEE Lausanne Power Tech, 1090–1093. doi: 10.1109/pct.2007.4538467
  17. Qiang, Z., Ben, D. W., Bin, H., Yong, P., Ming, L. R. (2011). Singular Spectrum Analysis and ARIMA Hybrid Model for Annual Runnoff Forecasting. Water Resources Management, 25 (11), 2683–2703. doi: 10.1007/s11269-011-9833-y
  18. Shchelkalin, V. N. (2012). Dekompozitsionnyy podkhod prognozirovaniya vremennykh ryadov na osnove metoda «Gusenitsa»-SSA. Materialy 14-y Mezhdunarodnoy nauchno-tekhnicheskoy konferentsii SAIT. Kiev, 258–259.
  19. Shchelkalin, V. N. (2011). Ot idei metodov «Gusenitsa»-SSA i Boksa-Dzhenkinsa do dekompozitsionnogo metoda prognozirovaniya i dekompozitsionnoi INS. Vostochno-Evropeiskii zhurnal peredovykh tekhnologii, 59–69.
  20. de Souza, R. M., Menezes, M., Souza, R. C., Pessanha, J. F. M., Junior, L. A. T., Cassiano, K. M. (2012). SSA, Hierarchical Clustering and ARIMA Models in Forecasting Reservoir Inflows. The 32nd Annual International Symposium on Forecasting ISF, 11–25.
  21. de Souza, R. M., Menezes, M., Souza, R. C., Pessanha, J. F. M., Junior, L. A. T., Cassiano, K. M. (2012). Streamflow Forecasting Using Singular Spectrum Analysis, ARIMA Model And Artificial Neural Network. The 32nd Annual International Symposium on Forecasting ISF, 33–47.
  22. Cassiano, K., Menezes, M., Junior L. F., Pessanha, J., Souza Raf., Souza R. (2014). ARIMA Model, Neural Networks and SSA in the Short Term Electric Load Forecast. Economic Forecasting – Past, Present and Future: International Symposium of Forecasting ISF, 1–23.
  23. Boks, Dzh., Dzhenkins G. (1974). Analiz vremennykh ryadov. Prognoz i upravlenie. Mir, 406.
  24. Evdokimov, A. G., Tevyashev A. D. Ob odnom algoritme vyyavleniya skrytoi periodichnosti. Avtomatizirovannye sistemy upravleniya i pribory avtomatiki. Khar'kov, 43, 60–64.
  25. Golyandina, N. E. (2004). Metod «Gusenitsa»-SSA: prognoz vremennykh ryadov. Sankt-Peterburg: S.-Peterburgskiy gosudarstvennyy universitet, 52.
  26. Sedov, A. V. (2010). Modelirovanie obyektov s diskretno-raspredelennymi parametrami: dekompozitsionnyy podkhod. Moskva: Nauka, 438.
  27. Chumak, O. V. (2011). Entropii i fraktaly v analize dannykh. Izhevsk: Institut komp'yuternykh issledovaniy, 164.

Published

2014-12-22

How to Cite

Щелкалин, В. Н. (2014). Hybrid mathematical models and methods of time series forecasting taking into account external factors. Eastern-European Journal of Enterprise Technologies, 6(4(72), 38–58. https://doi.org/10.15587/1729-4061.2014.31729

Issue

Section

Mathematics and Cybernetics - applied aspects