Data Analysis Methods for Support Decision Making at Management of Complex Systems

Abstract

Problems of organizing decision support in the management of complex systems based on data mining are discussed in this article; also, the state of the art is presented. The specificity of data used for analysis is noted, along with the array of historical and current data. The problem is stated, methods for organizing decision-making information support with set of recommendations are proposed. The proposed methodology includes collection and preparation of data for analysis, identification of new knowledge based on similarity of objects using clustering, their integration with expert knowledge, formalization of knowledge and formation of the knowledge base, obtaining solutions while making use of knowledge and inference engine. Tools for data mining, namely, the analytical platform Deductor Studio, are shown. The results of experimental studies based on the proposed method are provided. The system of production rules and the inference engine are proposed to use for organization of decision-making support. In this case, a consequent set of rules is presented in the form of recommendations for opening branch, which is shown as the boundary values of several characteristics. The boundary values are determined by results of an object entering the cluster, which is revealed by conducting cluster analysis using neural network apparatus. The specially developed software solution is used to implement solutions based on the system of production rules. Keywords—Data mining, production rules, Kohonen maps/Self-organizing map, inference engine, decision support.