Babii A. Models, methods and intelligent information technology for analysis of heterogeneous sequences

The object of this thesis research is the process of analisys data of heterogeneous sequences to assess the current state of the domain. Purpose of the thesis research is solution of the actual scientific and practical problem of the development of models, methods and intelligent information technology for the analysis of heterogeneous data sequences to assess the current state of the domain for information-analytical system. Research methods: methods of fuzzy set theory, regression analysis, fuzzy approximation of data, mathematical statistics.The method is first proposed in the thesis research to determine significant factors of the fuzzy regression model of heterogeneous data; this method, in contrast to the actual methods, includes stages of factors selection in accordance with the criterion of equal significance of angles of deviation between the vector of errors and vectors of variables, as well as selection of subsets of significant factors with coefficients that exceed the threshold value, that allows avoiding any over-fit of the fuzzy linear regression and receiving the subset of significant factors on the basis of the finite number of iterations.The development of the trend-seasonal model of heterogeneous sequences has gone further; in contrast to actual models its trend component is given in the form of interpolated averaged values with regard to membership function, which is associated with every fuzzy class that permits to use this model for short series without any loss of the boundary data and, therefore, improve the modeling accuracy. The method of filtration of components of heterogeneous time sequences received its further development, in contrast to actual methods the initial sequence is broken into a finite number of fuzzy segments in order to find out the trend, and for each of these segments the averaged value is calculated with regard to the membership function associated with the fuzzy segment, that permits to take into account the boundary values of the series in order to find the trend component. Intellectual information technology for the analysis of heterogeneous data sequences has been improved, which, unlike the existing ones, contains methods and means for determining the meaningful factors of a fuzzy regression model and filtering components of time series of data, which allows to improve the efficiency of the analysis of heterogeneous sequences. Implemented information technology were used in the activities of LLC "Endevour", the Main Directorate of the National Police of the Kharkiv region. The proposed methods are used in the educational process of the Kharkiv National University of Radio Electronics and can be applied for data analysis in the activities of law enforcement agencies, medical organizations and other institutions.