Байєсівський аналіз даних у моделюванні та прогнозуванні нелінійних нестаціонарних фінансово-економічних процесів

• Datum: 2023
• Hauptverfasser: Trofymchuk, Oleksandr, Bidyuk, Petro, Prosyankina-Zharova , Tetyana, Terentiev, Oleksandr
• Format: Artikel
• Sprache: English
• Veröffentlicht: V.M. Glushkov Institute of Cybernetics of NAS of Ukraine 2023
• Schlagworte: нелінійні нестаціонарні процеси; байєсівські методи; моделювання; прогнозування; узагальнені лінійні моделі
• Online Zugang: https://jais.net.ua/index.php/files/article/view/114
• Назва журналу: Problems of Control and Informatics

The study focuses on some aspects of modeling and forecasting the nonlinear nonstationary processes (NNP) of applying the modern Bayesian methods of data, in particular, generalized linear model (GLM) that are popular in analysis of NNP. All Bayesian techniques of data analysis are very popular today thanks to their flexibility, high quality of results, availability of possibilities for structural and parametric optimization and adaptation to new data and conditions of functioning. The structural and parametric adaptation of Bayesian generalized linear models supposes taking into consideration the following elements: number of equations that are necessary for adequate formal description of the processes under study; availability of nonlinearity and nonstationarity; type of random disturbance — its probability distribution and corresponding parameters; order of model equations, and some other structural elements. Such approach to modeling improves model adequacy and quality of final result of their application. Parameter estimation of the models can be performed by making use of rather wide set of methods, more precisely the following: ordinary LS (OLS), nonlinear LS (NLS), maximum likelihood (ML), the method of additional variable (MAV), and Monte Carlo for Markov Chain (MCMC). The last method is distinguished by universality of application to estimation of linear and nonlinear models. Besides, each of Bayesian approaches to data analysis is well supported by appropriate sets of statistical criteria that make it possible thorough quality analysis of intermediate and final results of computations. Illustrative examples are presented the usage of the Bayesian approach for analysis and forecasting of NNP, in particular, in specialized intellectual decision support system.