: ..
: PageRank
: 111
:
: 2024
: .. PageRank // . 111. .: , 2024. .81-96. DOI: https://doi.org/10.25728/ubs.2024.111.3
: PageRank, ,
(.): PageRank, centrality, Markov chains, weighted directed graph
: . , , . PageRank . . , PageRank, . . , . , . , . .
(.): The work is devoted to finding the centrality of nodes of weighted graphs. The relevance of this task is due to the fact that ignoring the weights of the arcs of the graph when finding the centrality of its nodes is unacceptable for a number of applied tasks, primarily related to tasks from the financial sphere. In the classical formulation of the PageRank algorithm, part of the information about the weights of connections is lost when forming a matrix of transient probabilities from the adjacency matrix. This effect has been demonstrated in this article. A method for determining the centrality of network nodes is proposed, based on the PageRank algorithm, which allows taking into account all the weights of the links. The graph of financial transactions was considered as an example. The nodes of the graph are the clients of a commercial bank, including the bank itself, and the arcs are money transfers between nodes. The ranking quality was determined by comparing various centrality measures with an external node parameter unrelated to the network characteristics of the transaction graph. According to the results of the study, it was shown that the proposed centrality measure ranks the most important nodes of the graph in the best way compared to other centrality measures. The convergence of the proposed algorithm was also demonstrated.
PDF
: 105, : 27, : 8.