Time Series Modeling with Genetic Programming Relative to ARIMA Models

Created by W.Langdon from gp-bibliography.bib Revision:1.3973

@InProceedings{klucik:2009:NTTS,
  author =       "Miroslav Klucik and Jana Juriova and Marian Klucik",
  title =        "Time Series Modeling with Genetic Programming Relative
                 to ARIMA Models",
  booktitle =    "New Techniques and Technologies in Statistics, NTTS
                 2009",
  year =         "2009",
  address =      "Brussels, Belgium",
  month =        "18-20 " # feb,
  organisation = "EUROSTAT and European Commission",
  keywords =     "genetic algorithms, genetic programming, symbolic
                 regression, ARIMA",
  URL =          "http://epp.eurostat.ec.europa.eu/portal/page/portal/research_methodology/documents/POSTER_4P_TIME_SERIES_MODELLING_KLUCIK.pdf",
  size =         "10 pages",
  abstract =     "INFOSTAT, the research institution of the Statistical
                 Office of the Slovak Republic, is intending to
                 supplement its model tools (ECM, ARIMA) with modern
                 heuristic methods for analyzing and forecasting the
                 macroeconomic reality. The initial research is
                 concentrated on time series modelling using genetic
                 programming and comparing the results with a more
                 conventional ARIMA model. Genetic programming tool
                 based on evolutionary computation technique can find
                 not only optimal parameters of a searched function but
                 also its structure. Our experiments deal with modeling
                 and forecasting of the industrial production for
                 Slovakia and European Monetary Union. For our purpose
                 the genetic programming tool is kept as simple as
                 possible. The predicted variables are estimated by the
                 concept of symbolic regression. The solutions of
                 symbolic regression are expressed in a tree-type
                 structure. Concerning the ARIMA approach, we have used
                 seasonal ARIMA models that satisfied all the quality
                 model conditions. Both methods' performance was tested
                 in a twelve-month forecasting. The second experiment
                 involves the simulation of shocks for each model. The
                 GP model manages to compete with ARIMA models in all
                 cases. Finally we show a way to depict a complicated
                 nonlinear solution in a simply understandable form. The
                 continually changing and hardly predictable environment
                 of contemporary and future global economy will require
                 a multidisciplinary approach to approximate the complex
                 reality. The GP instrument with its flexibility and
                 efficiency manages to confront these challenges with
                 promising results.",
}

Genetic Programming entries for Miroslav Klucik Jana Juriova Marian Klucik

Citations