THE INTERVAL FORECASTING OF DYNAMIC INDICATORS BASED ON LOGISTIC REGRESSION MODELS

Receipt date: 
16.09.2017
Year: 
2017
Journal number: 
УДК: 
519.688
DOI: 

10.26731/1813-9108.2017.4(56).122-131

Article File: 
Pages: 
122
131
Abstract: 

Formalization, software implementation and accuracy testing of interval forecasting of real dynamic indicators with different statistical properties using logistic regression methods with and without regularization have been carried out in the paper. The interval forecasting of dynamic indicators involves determining that their future values belong to the preset intervals based on the probability estimates. Since we do not estimate a future value of the indicator, but rather the interval in which it is going to be located, we have called this forecasting method «interval forecasting». For accuracy testing of interval forecasting, we have used real dynamic indicators with different statistical properties which depend on the stationarity of location and scale parameters. For testing of stationarity of location parameter value with the passage of time, we have used a modified Kruskal-Wallis test, and for testing of stationarity of scale parameter value with the passage of time, we have used a modified Fligner-Killeen test. The results showed that in most cases for dynamic indicators with different statistical properties, the logistic regression model without regularization has demonstrated the best interval forecasting accuracy. Thus, in practice, we recommend the model of the logistic regression without regularization for the interval forecasting of real dynamic indicators, in particular, to construct forecasting ensembles.

List of references: 
  1. 1.   Mitrea C.A. A Comparison between neural networks and traditional forecasting methods: a case study. International journal of en-gineering business management, 2009, No. 1(2), pp. 19–24.

    2.   Gooijer J.G., Hyndman R.J. 25 years of time series forecasting. International journal of forecasting, 2006, No. 22 (3), pp. 443–473.

    3.   Shumway R.H. Time series analysis and its applications with R examples. Springer, 2011, 609 p.

    4.   Wang H., Li G., Wang H, Deep learning based ensemble approach for probabilistic wind power forecasting. Applied energy. 2017, No. 188, pp. 56–70.

    5.   Vernay M., Lafaysse M., Merindol L., Ensemble forecasting of snowpack conditions and avalanche hazard. Cold regions science and technology, 2015, No. 120, pp. 251–262.

    6.   Elliot G. Predicting binary outcomes [Electronic resource]. URL: http://econweb,ucsd,edu/~grelliott/BinPred,pdf (access date: 22.08.2017).

    7.   Yoder M., Cering A.S., Navidi W.C. Short-term forecasting of categorical changes in wind power with Markov chain models. Wind energy, 2014, No. 17, pp. 1425–1439.

    8.   Lahiri K., Yang L. Forecasting binary outcomes. Handbook of economic forecasting [Electronic resource]. URL: http://www.albany.edu/economics/research/workingp/2012/lahiriyang,pdf (access date:12.05.2017).

    9.   Krakovskii Yu.M., Luzgin A.N. Algoritm interval'nogo prognozirovaniya dinamicheskikh pokazatelei na osnove robastnoi veroyatnostnoi klasternoi modeli : elektron. zhurn. [Algorithm for interval forecasting of dynamic indicators based on a robust probabilistic cluster model: electron. journal.]. Nauka i obrazovanie, 2016, No. 11, pp. 113–126. URL: http://technomag.neicon.ru/doc/849839.html (access date: 01.07.2017).

    10. Murata A., Fujii Y., Naitoh K. Multinomial logistic regression model for predicting driver's drowsiness using behavioral measures. Procedia manufacturing, 2015, No. 3, pp. 2426–2433.

    11. Arbues F. Determinants of behavior toward selective collection of batteries in Spain, A bivariate probit model. Resources conservation and recycling, 2016, No. 106, pp. 1–8.

    12. Cui M., Ke D., Sun Y. Wind power ramp event forecasting using a stochastic scenario generation method. IEEE Transactions on sustainable energy, 2015, No. 6, pp. 422–433.

    13. Dreiseitl S., Ohno-Machado L. Logistic regression and artificial neural network classification models: a methodology review. Journal of biomedical informatics, 2002, No. 35, pp. 352–359.

    14. Kliestik T., Kocisova K., Misankova M. Logit and probit Model used for prediction of financial health of company. Procedia eco-nomics and finance, 2015, No. 23, pp. 850–855.

    15. Krakovskii Yu.M., Luzgin A.N. Interval'noe prognozirovanie nestatsionarnykh dinamicheskikh pokazatelei na osnove modeli veroyatnostnoi neironnoi seti [Interval prediction of non-stationary dynamic indicators based on the probabilistic neural network model]. Nauchnaya mysl' [Scientific thought], 2016, No. 1, pp. 116–122.

    16. Krakovskii Yu.M., Luzgin A.N. Proverka nestatsionarnosti dinamicheskikh pokazatelei po kriteriyu sdviga Kraskela-Uollisa [Checking the nonstationarity of dynamic indicators by Kruskal-Wallis shift criterion]. Baikal'skii Vestnik DAAD [Baikal Letter DAAD], 2016, No. 1, pp. 17–23.

    17. Kobzar' A.I. Prikladnaya matematicheskaya statistika [Applied mathematical statistics]. Moscow: Fizmatlit Publ., 2006, 816 p.

    18. Hettmansperger T.P., McKean J.W. Robust nonparametric statistical methods. New York : Chapman-Hall. 2011, 553 p.

    19. Kloke J., McKean J.W.  Nonparametric statistical methods using R. New York : Chapman-Hall, 2014, 283 p.

    20. CS229 Lecture notes [Electronic resource]. Stanford University. URL: https://see.stanford.edu/materials/aimlcs229/cs229-notes1.pdf (access date: 10.09.2017).

    21. Minka T.P. Algorithm for maximum-likelihood logistic regression [Electronic resource]. URL: https://tminka.github.io/papers/logreg/minka-logreg.pdf (access date: 08.07.2017).

    22. Genkin A., Lewis D.D., Madigan D.D. Sparse logistic regression for text categorization [Electronic resource]. URL: http://www.ics.uci.edu/textasciitilde smyth/courses/cs277/papers/genkin_logistic_regression_sparse.pdf  (access date: 02.02.2017).

    23. The R-Project of statistical computing [Electronic resource]. URL: http://www.r-project.org (access date: 10.11.2017).

    24. Krakovskii Yu.M. Programmnoe obespechenie interval'nogo prognozirovaniya nestatsionarnykh dinamicheskikh pokazatelei [Software for interval forecasting of non-stationary dynamic indicators]. Vestnik IrGTU [Proceedings of Irkutsk State Technical University], 2015, Vol. 1, No. 4, pp. 12–16.

    25. RStudio [Elektronnyi resurs]. URL: https://www.rstudio.com (data obrashcheniya: 11.11.2016).

    26. Package ‘gWidgets2’ [Electronic resource]. URL: https://cran.r-project.org/web/packages/gWidgets2 (access date:  01.01.2017).

    27. RGtk2. [Electronic resource]. URL: https://cran.r-project.org/web/packages/RGtk2 (access date: 02.15.2017).

    28. A Library for Large Linear Classification [Electronic resource]. URL: https://www.csie.ntu.edu.tw/\textasciitilde cjlin/liblinear (access date: 15.05.2017).

    29. Rufibach K. Use of Brier score to assess binary predictions. Journal of clinical epidemiology, 2010, No. 63(8), pp. 938–939.

    30. Air Quality Data of Switzerland [Electronic resource]. URL: https://cran.r-project.org/web/packages/SwissAir/index.html (access date: 11.03.2016).

    31. DataMarket [Electronic resource]. URL: https://datamarket.com/data (data obrashcheniya: 10.04.2017).

    32. SRCP [Electronic resource]. URL: http://www.crsp.com/products/documentation/crsp-calculations (access date: 11.03.2017).

    33. Madras Monthly Sea Level, CRU [Electronic resource]. URL: http://www.comp-engine.org/timeseries/time-series_data/data-11114 (access date: 18.05.2017).

    .2017).