Kobylin I. Fuzzy clustering of time series in data stream mining

In dissertation work the new actual task of unclear clusterization of sentinel rows is untied in the intellectual analysis of streams given. The object of study is the process of intelligently analyzing the flow of data in the form of time series. A fuzzy clustering method is proposed that works effectively under conditions of intersection of classes that are not subject to the effect of concentration of norms and works online with asynchronous non-uniformly quantized time series through the use of a special objective function. A sequential online clustering method for multidimensional time series is proposed, based on the apparatus of hybrid systems of computational intelligence, which made it possible to solve the problem of clustering data that are sequentially received for processing with non-uniform quantization cycles. An adaptive probabilistic and probabilistic clustering method was developed based on a special type metric based on the analysis of the tangents of the slopes of the time series, which made it possible to simplify the numerical implementation of the procedure and solve the clustering problem for unevenly quantized time series. Model of robust adaptive identification of non-stationary time series in online flow of data flow шіare presented, which are characterized by simplicity of computational implementation and resistance to anomalous emissions. The worked out methods were programmatic realized and used for the row of practical introductions – in particular in the tasks of monitoring of medical data in on-line mode.