Bidyuk P.I.

Analysis of Investment and Socio-Economic Indicators Using the Methods of Modeling for the Limited Historical Data Sets

In this paper, we analyze the development of economic region of Ukraine using statistical data. Specific features of mathematical models construction techniques for analysis and short- and medium term forecasting of regional macroeconomic processes are considered. We study the approaches to constructing mathematical models with short time series and using intellectual data analysis techniques such as principal component method, Bayesian networks, and regression with lagged variables, extended autoregression and trend polynomials.

Data Analysis with Bayesian Models

In this paper, we propose to review some Bayesian data analysis models, namely the models with one and several parameters. Using statistical data and expert estimates, we develop the methodology for probabilistic models construction in the form of Bayesian networks. The methodology enables us to construct high adequacy probabilistic models for solving classification and forecasting problems. Furthermore, we propose an integrated dynamic network model by combining probabilistic and regression approaches.

Informational Decision Support System for Forecasting Financial and Economic Processes Based on Structural and Parametric Adaptation of Models

A concept is proposed for solving the problem of adaptive forecasting that is based on the system analysis methodology and combined use of preliminary data processing techniques, mathematical and statistical modeling, forecasting and optimal state estimation of the processes under study. The cyclical adaptation of a structure and model parameters on the basis of a set of statistical characteristics of a process under study provides a possibility for reaching high quality estimates of forecasts with condition that data is informative.

Diagnosing Technical Objects Based on Artificial Immune Systems and Bayesian Networks

A generalized information technology for building artificial immune systems for solving problems of technical diagnosis is proposed. The technology allows creating a mathematical description of the drift parameters and detecting anomalies in the operation of complex technical systems. The novel method and algorithm for detecting the location and type of failure of complex engineering system with Bayesian networks and information content criteria are created. The information technology for neural networks development based on the theory of immune systems is developed.

A modified Bayes method of forming network structure

The present paper offers a modified procedure for learning Bayesian network in terms of the statistical data. Moreover, the introduced recursive formulas enable the computational process optimization, and the implemented algorithm of heuristic search for discovering BN structure allows avoiding dependence of vertices from their initial ordering.

The methods of constructing dynamic Bayesian networks

This study proposes a method of constructing dynamic Bayesian networks based on statistical data. It consists of two phases: building a static network structure and dynamic structure of the network, which determines the relations between two neighboring time intervals. The structure obtained is used to form the output at each time interval. Finally, we illustrate how this method can be applied to actual data

Adaptive forecasting of financial and economic processes based on the principles of system analysis

Based on the principles of system analysis, we propose the concept of developing and implementing of adaptive forecasting systems. It allows taking into account various types of uncertainties and increasing the quality of forecast estimates. The forecasting system includes two adaptation loops, whose functioning aims at boosting the model quality and forecast estimates. We provide an example of this system application.

The Comparative Analysis of the Models’ Characteristics for Credit Rating

This paper analyzes basic methods of the credit rating (decision tree and logistic regression). More precisely, we propose to use the models on Bayesian networks. We construct the appropriate models for the decision tree, logistic regression and Bayesian networks. We calculate the common accuracy of each model, I-type and II-type errors, and build ROC-curves and GINI-index. Finally, we arrive at the conclusion that it’s reasonable to use Bayesian networks.

Modelling of heteroscedastic processes of transition economy using alternative methods of models estimation

This paper deals with modelling of heteroscedastic processes of the transition economy. By employing linear and nonlinear methods, notably the least squares method, Marquardt method, and Monte-Carlo method, we estimate the coefficients. Perspective taking, this problem solving seems to be quite urgent and promising for successful estimation of inflation, production costs modelling, as well as forecasting prices on the stock exchange and other heteroscedastic processes.

Decision support system based for analysis of financial data

In this paper, the basic features of financial analysis of an enterprise are considered, and the technology of the financial data analysis based on the integrated approach is proposed. We also show that this technology can be applied by giving the examples of sales analysis and performing the future forecasts as well as estimating defaults of credit borrowers. Moreover, we highlight the architecture of the proposed information system of decisionmaking support for the commercial bank.