Methodology of the Extreme Value Theory Application for Data Analysis
DOI:
https://doi.org/10.20535/1810-0546.2016.1.64895Keywords:
Extreme value theory, Maximum likelihood estimator, Simulation and modeling, Decision support systemAbstract
Background. To solve the problems of modeling and forecasting on the basis of large datasets (including singular ones) in conditions of uncertainty it is necessary to develop integrated information decision support systems (DSS). A methodology is proposed for application of extreme value theory for statistical models development and DSS on their basis.
Objective. The goal of the study is in application of the extreme value theory for analysis and estimation of model parameters on the basis of random samples. It is necessary to develop an effective methodology for analysis of pseudorandom sequences and estimation of unknown model parameters; to present examples of analysis using extreme value theory and software developed.
Methods. To solve the problems stated the following approaches were used: pseudorandom sequences generating procedures; probabilistic distributions of the extreme value theory, and methods for estimating unknown model parameters. A multistep methodology is proposed for extreme values processing and DSS is developed for analysis and modeling of pseudorandom sequences.
Results. Using the DSS developed and generated statistical data as well as proposed methodology the procedure was developed for extreme values analysis. The procedure is to be used for estimating of forecasting models for the process of various origin. A comparative analysis of parameter characteristics for GEV-distributions was performed.
Conclusions. Using the instrumentation developed it was shown that the proposed methodology for processing extreme values is convenient for analysis of singular datasets. This is substantiated with the high quality approximation of theoretical probability density by empirical curve. A comparison of model parameters estimation results showed that the estimates converge faster when parameters of form and scale are defined more exactly.References
S. Coles, An Introduction to Statistical Modeling of ExtremeValues. London, UK: Springer-Verlag, 2009, 209 p.
R.L. Smith, Extreme Value Theory. Chapel Hill: University of North Carolina, 2009. – 178 р.
M. Kuhn and K. Johnson, Applied Predictive Modeling. New York: Springer, 2013, 600 p.
R.H. Shumway and D.S. Stoffer, Time Series Analysis and its Applications. New York: Springer, 2006, 598 p.
A. Romano and G. Secundo, Dynamic Learning Methods. NewYork: Springer, 2009, 190 p.
P. McCullagh and J. Nelder, Generalized Linear Models. New York: Chapman & Hall, 1989, 526 р.
R.S. Tsay, Analysis of Financial Time Series. New Jersey: John Wiley & Sons, Inc., 2010,715 p.
W.R. Gilks et al., Markov Chain Monte Carlo in Practice.Boca Raton: Chapman and Hall/CRC, 2000, 486 p.
P.I. Bidyuk and L.O. Korshevnyuk, Design of Decision Support Systems. Kyiv, Ukraine: NTUU KPI, 2010, 340 p. (in Ukrainian).
C.W. Holsapple and A.B. Whinston, Decision Support Systems. New York: West Publishing Company, 1994, 860 p.
A.M. Law and W.D. Kelton, Simulation Modeling and Analysis. New York, McGraw Hill, 2000, 760 p.
J. Beirlant, Statistics of Extremes: Theory and Application. New York: John Wiley & Sons, Inc., 2004, 505 p.
P.I. Bidyuk and S.V. Trukhan, “Estimation of generalized linear models using Bayesian approach in actuarial modeling”, Naukovi Visti NTUU KPI, no. 6, pp. 49–55, 2014 (in Ukrainian).
Downloads
Published
Issue
Section
License
Copyright (c) 2017 NTUU KPI Authors who publish with this journal agree to the following terms:- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under CC BY 4.0 that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work
