“Soft” calculation methods in the evaluation objects of complex systems

Abstract

Today, humanity in its activities quite often interacts with complex systems of economic, transport, construction and many other industries. The complexity of these systems will be manifested in a large number of connections between elements that are connected not only to each other, but also to other subsystems. Each of these industries consists of already well-studied and analyzed systems, such as employee payroll. But, in addition, they also include systems based on a qualitative component that do not yet have a developed mathematical description. Such systems include the influence of the internal climate that unites the members of the organization's team on labor productivity. Scientists have proposed many different approaches to solving this problem based on the use of statistical, differential methods. Even machine learning, which is quite popular today, is also used in these tasks. But the vast majority of them have a complex structure, which is manifested in the use of the apparatus of higher mathematics. Because of this understanding of the model itself, its application recedes into the background. Accordingly, the first place is the requirement to know and navigate in a complex mathematical description. Because of this, only a narrow circle of specialists is able to use models built using this technology. The authors of this article propose their approach, which is based on the method of artificial intelligence. We are talking about "soft" methods consisting of such components as neural networks, genetic algorithms and fuzzy sets. It was on the latter that the authors focused most of their attention for evaluating the objects of complex systems. Of course, one method is not enough for the developed model to adequately represent the operation of the system under study. And thus, to ensure the possibility of its dynamic description, genetic algorithm methods were also used. Of course, these methods also have a mathematical description. But, in contrast to strict mathematical methods, in these two approaches of artificial intelligence, the visual component is quite well represented. This allows you to almost immediately answer the question of how this or that value was obtained during the operation of the model, with the option of not using formulas for this. As a result of the work carried out, a structural fuzzy model was created, which was expanded by the methods of crossing over and selection.