Methods of clustering indicators in the financial assessment of interrelated counterparties

Abstract

The article examines the main approaches to clustering of similar data characterizing the economic activity of these entities using fifty similar counterparties as an example. A comparison is made between the basic clustering algorithms using the Euclidean distance and the typical agglomerative algorithm based on the Mahalanobis distance, which considers the covariance estimates of elements from the set under consideration. The article examines the consistent application of different types of algorithms to economic entities. In addition, it examines the change in dependence when applying various analytical procedures. Based on the analysis, conclusions are made about the advantages and disadvantages of the above algorithms, and it is noted that it is advisable to consistently apply algorithms of various types to identify hidden causes that affect the activities of legal entities and that are not obvious when conducting a standard financial analysis of the activities of business entities.