Sydorova M. Information technology of cluster analysis of multivariate time series in the processing of the monitoring results.

The object of research is the process of samples monitoring by the set of features. The aim of the thesis is the development of information technology of cluster analysis of multivariate time series for definition groups of objects by similarity of attributes as in every moment and in the time period of observation, evaluating the results and supporting decision making in the processing of monitoring data. Research methods - methods of cluster analysis, classification, probabilistic and statistical analysis, decision theory, time series analysis, computational mathematics, programming technology. In this paper, the method of cluster analysis of multivariate time series, that allows to dynamically obtaining cluster structure with new observation results, using prior information about the similarity of objects, has been designed. A new similarity measure of time series, which takes into account both proximity values and the degree of connection comparable features, has been proposed. The information technology of multicriteria evaluation and improve the stability of clustering results has been proposed to determine the best partition of objects. The system of mining "Medisa" that based on the proposed methods and techniques for solving the cluster analysis, classification, prediction, statistical analysis and visualization of monitoring data have been created. The results of the thesis are introduced into the activities of 5 state enterprises and institutions. Scope of application - information technology, processing of the results of monitoring, cluster analysis of multivariate data and time series.