Algorithm of Hierarchical Structure Integral Indicators Construction Based on Define Matrices

The article is devoted to the problem of optimal design of hierarchical structure integral indicator based on hierarchical analysis of forward and backward movement and methods of automatic clustering and ranking. Today, there are a significant number of methods for constructing criterion importance vectors by a group of experts; the problem of integrated application of system analysis methods of hierarchical structures is widely represented in scientific research. However, the problems of identifying elements in the hierarchy, determining the contribution of elements to achieving the goal of the system, assessing the adequacy of the resulting hierarchical decomposition, and maintaining the relevance of this representation remain unresolved. To increase the speed of decision support information systemoperation in hierarchical structures, one may use historical knowledge about the results of already made decisions. Utilizing solutions that have worked well in the past allows narrowing down the solution range. This reduces the complexity of decision-making problems. Using information about the level of systemintegral indicators under study allows making decisions based on monitoring all states of the constituent elements. The article proposes to use an original algorithm for constructing integral indicators of hierarchical structures. It is based on the hierarchy analysis method and computerized multidimensional data analysis. In contrast to T. Saaty’s method of analyzing hierarchies, when forming a general expert assessment, it is recommended to use the average deviation of the consistency ratios of the expert’s pairwise comparison matrices from the average for the expertgroup as the expert’s weight. Refining the weights and structure of the integral indicator using the given algorithm allows for manual and automatic parameteraggregation into clusters of a hierarchical structure together, and with mutual horizontal influence of hierarchy elements, combining them into a cluster with common weight formed in a fuzzy logic system.