Мережі, що визначаються динамікою тематичних інформаційних потоків

A technique for the formation, clustering and visualization of the so-called correlation networks is proposed. Links between nodes in such networks correspond to correlation values between vectors - sets of parameters corresponding to these nodes. To build network structures for each node (subject), vectors - arrays of numbers corresponding to thematic documentary collections are formed. For this, it is planned to use a content monitoring system for social media. When processing thematic queries, many vectors of dynamics are determined. These vectors correspond to the given topics/entities. After the formation of these vectors, a correlation network is formed. This network can be considered as a way to preserve and visualize entities that are objectively interconnected. After this, the set of maximum cross-correlations between the obtained vectors is calculated, the corresponding adjacency matrix is formed, and this matrix in CSV format is saved. Further, the formed matrix is transmitted for processing and visualization to the Gephi network structure analysis system. After that, the modularity classes of objects and subsequent clustering are determined for the network. Network visualization is also performed in the Gephi system. The above approach, in contrast to the existing ones, has such advantages as a relatively low dimension of parameter vectors corresponding to the topics; independence from the language of the documents — the parameter vectors are determined only by requests to the content monitoring system, which can contain words in different languages; relative ease of implementation. The above technique can be used in information and analytical systems for various purposes for the analysis of arrays of entities without explicit relationships between them. Correlation networks can be considered as the basis for the construction of probability networks and the use of fuzzy semantic network technologies for further scenario analysis. Tabl.: 1. Fig.: 4. Refs: 6 titles.