Full-texts publications of the department - Part 2

Full-texts publications of the department - Part 2

M. Didkovska, A. Gogolev. Comparative analysis of clustering algorithms

A constant information production rate’s growth leads to disproportionality of "information noise" due to the weak data structuring, inconsistency of formally relevant information and its multiple duplication. Search in systematically updated information space can be simplified by means of categorization. The problem of "information noise" is important for online stores. Their databases contain about half a million products. Prices are constantly changing, some goods are no longer for sale, new items appear. For this reason, online store has to update constantly their databases. And every time during update items must be categorized. Currently there is no unified approach to this problem. In the article the mathematical formulation of goods’ categorization problem is represented, the following stages are pointed out: indexing, classification and evaluation. Experimental study of classifiers (naive Bayes classifier, SVM method and decision tree) has shown that SVM method is the most effective to solve the problem of categorization.

M. Didkovska, V. Naryzhnyi Features of constructing a multimedia database management system

For today information is one of the most valuable resources, so the requirements to the systems of storage have to be severe. With the invention of high speed data transfer, the traditional text information is complemented by multimedia: photo, audio and video materials. Nowadays, there is no unified approach to the architecture of multimedia databases, multimedia data require more complex functionality then processing of textual information. In this work the basic requirements for multimedia database were specified. The functions and approaches used for work with multimedia objects were pointed out. Two approaches to the organization of multimedia databases were considered – with usage of file servers and without. The experimental study has shown the practicability of file servers’ using for storage of multimedia information, because this approach provides higher speed of data read-write

M. Didkovska, J. Bukhtiyarov, D. Gorobchenko The system of automated tests’ generation based on UML Use Case diagrams

System for automated tests’ creation, that allows to reduce the time and costs required for the testing process, is proposed. The results of system functionality are tests’ templates that contain input data, conditions of execution and expected results. Tests’ creation in automatic mode based on the analysis of the UML use case diagrams allows to avoid omission of important tests and the incomplete coverage of specifications. The proposed structure of the automated tests’ generation system may be recommended for use in the further research, in order to generate tests not only from the use case diagrams, but from others UML diagrams such as State-Chart diagram and Sequence diagram.

M. Didkovska, I. Bogomolov. Analysis of software quality problem

This paper is focused on qualitative software characteristics, what determines software quality and methods to ensure and improve it. In the research the problem was investigated from two points of view: software development and project management. Concerning software development, was formulated what properties have qualitative software and compile the best practices to ensure and improve these properties and overall software quality. Concerning project management, current theory was analysed and found another value – development team, which significantly influences quality. Also, dependency between team productivity and number of team members was stated.

Bidyuk P.I. Adaptive forecasting of financial and economic processes on the principles of system analysis

Concept of development and implementation of adaptive forecasting systems is proposed that is based on the system analysis principles. The approach proposed provides a possibility for taking into consideration the uncertainties of various types and to increase the quality of forecast estimates. The forecasting system includes two adaptation loops the functioning of which is directed towards improvement of model quality and forecast estimates respectively. An example of the system application is given.

Bidyuk P.I., Litvinenko V.I., Gasanov A.S. Immune network based method for identification of turbine engine surging

A modified method and algorithm of negative selection has been developed to solve the problem of anomalies detection in functioning of complex engineering systems. The method and algorithm for anomalies identification uses the mechanisms of artificial immune networks. The distinctive feature of the algorithm is in updating of training process thanks to which the possibility of adaptive selection is implemented. The experimental study has shown high efficiency of the offered algorithm.

Lytvynenko V.I., Bidjuk P.I., Rogalsky A.F., Fefelov A.A., Fedchuk V.A. Structural and Parametric Synthesis of predictive RBF Neural Networks using Artificial Immune Systems

In this paper we describe the use of clonal selection algorithm for the synthesis of radial-basis networks for solving the problem of time series prediction

Bidyuk P.I., Terentyev A.N., Gasanov A.S. Constructing and Methods of Learning of Bayesian Networks

This analytical literature review discusses different methods under the general rubric of learning Bayesian networks from data. The basic concepts of Bayesian networks and their learning methods are introduced and reviewed. The methods are discussed for learning parameters of a probabilistic network, for learning the structure, and for learning hidden variables. Basic definitions and key concepts with appropriate illustrative examples are presented.

Bidyuk P.I., Terentyev A.N., Sverdel K.A. The construction of medical expert systems using Bayesian networks

Bayesian networks represent a useful and seriously claimed instrument for implementation in data-mining systems of various applications. In the article an analysis of existing methods of Bayesian networks construction and probability inference. Examples for using BN in medicine are presented.

Bidyuk P.I., Terentyev O.M. The methodology of constructing and application of Bayesian networks

Specific features of the problem of structure determining and Bayesian networks learning are considered. The method of a network constructing is proposed that is based on analysis of mutual information estimate between nodes as well as minimum description length approach. An algorithm of the method is applied to the known problem of “Asia” that includes 8 nodes. The computing experiments performed proved high effectiveness of the method proposed for constructing and learning the networks.

Zgurovsky M.Z., Bidyuk P.I., Terentyev O.M. The Method for adapting probabilistic Bayesian model to statistical data

A method for learning and adapting Bayesian network structure to new data is proposed that is based on application of Bayesian approach and K2 algorithm. The method provides additional possibilities for adapting the preliminary formed network structure to new data and saving computing time.

Bidyuk P.I. Application of the Monte Carlo's method for markov's chains to estimate the model of stochastic volatility

 

Pavluk O.V., Bidyuk P.I The methodology of constructing of dynamic Bayesian networks

 

Bondarenko V.G. Diffusion on a manifold of nonpositive curvature

We gives estimates of the fundamental solution of the parabolic equation on a manifold of nonpositive curvature that are independent of the dimention. As a corollary, sufficient conditions of equivalence for some class of measure in Hilbert space were established.

Bondarenko V.G. Сonstruction of the fundamental solution of disturbed parabolic equation

In present paper parabolic equation solution is build. The construction is reduced to iterative prosedure. And convergence of the latter is proven.