Structural and Parametric Adaptation of Probabilistic and Statistical Models for Financial Risks Assessment

Authors

DOI:

https://doi.org/10.20535/1810-0546.2018.3.131976

Keywords:

Structural-parametric adaptation, Probabilistic models, Financial risks, Criteria of decision quality

Abstract

Background. Uncertainties of various nature cause the emergence of financial risks that need to be evaluated in real time, taking into account a large set of explicit and implicit factors, adapting the model to the effect of random perturbations and changes in the environment. Sometimes the changes in the external environment can be so significant that the chosen model will be unacceptable for evaluation, and for this case it is necessary to develop a clear scheme of actions, a set of methods for the formation of candidate models, criteria for quality and the choice of the best, and to get the possibility of clarifying the type, structure and parameters of the model in dynamics.

Objective. Propose a method of structural and parametric adaptation based on probabilistic-statistical models, which will allow to evaluate financial risks through probability and possible losses, and take into account the limitations that arose already in the process of modeling.

Methods. Comprehensive application: optimal filter for pre-processing data and their preparation for model construction, regression model for formal description and prediction of conditional dispersion and probabilistic model in the form of Bayesian network for estimating the probability of possible losses.

Results. The proposed structural and parametric adaptation was used during modeling of various types of financial risks in the banking, telecommunication, investment and allowed to take into account the change of the environment by adapting mathematical models, reconfiguring their structure and changing their parameters in accordance with the imposed new requirements regarding the quality of the decisions made. As a result of computational experiments, it was found that the adaptation of the model as a reaction to changes in the cut-off threshold in processing loan applications allowed 17 % less errors in the wrong classification and thus reduced losses from unscrupulous borrowers by an average of 12 %.

Conclusions. The application of structural and parametric adaptation to predict the risks of different nature allows not only to choose the best mathematical model at the initial stage of risk assessment, but also to adapt it taking into account previous experience and the real work of decision support system, to specify the structure of the model according to the external disturbances, to adjust the parameters accordingly before changing the restrictions or imposing the new ones.

Author Biographies

Nataliia V. Kuznietsova, Igor Sikorsky Kyiv Polytechnic Institute

Наталія Володимирівна Кузнєцова

Petro I. Bidyuk, Igor Sikorsky Kyiv Polytechnic Institute

Петро Іванович Бідюк

References

V.A. Lukas, Theory of Management Technical Systems. Ekaterinburg, Russia: UGGU, 2015.

M.G. Popovich, Theory of Control Systems. Kyiv, Ukraine: Lybid, 2007.

P.I. Bidyuk et al., Analysis of Time Series. Kyiv, Ukraine: NTUU KPI, 2013.

A.J. McNeil et al., Quantitative Risk Management. Princeton: Princeton University Press, 2005, 538 p.

N.V. Kuznietsova and P.I. Bidyuk, “Scoring cards development for bank activities risks analysis”, Data Recording, Storage & Processing, vol. 19, no. 4, pp. 35–44, 2017.

M. Neil et al., “Using Bayesian networks to model expected and unexpected operational losses”, Risk Analysis, vol. 25, no. 4, pp. 963–972, 2005. doi: 10.1111/j.1539-6924.2005.00641.x

W.R. Gilks et al., Markov Chain Monte Carlo in Practice. New York, Chapman & Hall/CRC, 2000, 486 p.

K. Murphy. A Brief Introduction to Graphical Models and Bayesian Networks [Online]. Available: http://www.cs.ubc.ca/~murphyk/ Bayes/bayes.html

U. Kjærulff, “dHugin: A computational system for dynamic time-sliced Bayesian networks”, Int. J. Forecast., vol. 11, no. 1, pp. 89–111, 1995. doi: 10.1016/0169-2070(94)02003-8

N.V. Kuznietsova and P.I. Bidyuk, “Dynamic modelling of financial risks”, Inductive Modeling of Complex Systems, vol. 9, pp. 122–137, 2017.

Published

2018-06-05

Issue

Section

Art