Investigation of ways to accelerate the training of neural networks using genetic algorithms and the study of the dependence of the speed of genetic algorithms on the mutation rate. In this study, a program was implemented on the Unity graphics platform using genetic algorithms and mutations to determine their optimal coefficient. The experiment showed that the learning rate really depends on the mutation rate, and the highest learning rate was obtained at 5-7,5%.

Keywords: machine learning, deep learning, genetic algorithm, optimization, neural network, artificial neuron, mutation, artificial intelligence, non-player character, optimization

As a result of analyzing various factors that reduce the level of professional working capacity of the driver of passenger vehicles, a method of pre-trip computer assessment of his professional readiness has been developed. The software for personal computers and mobile devices has been developed, which allows to carry out express diagnostics of the functional states of the driver, based on test tasks and mathematical analysis of the obtained values.

Keywords: passenger driver, express diagnostics, pre-trip performance, functional status, testing methodology

This study describes the method of narrative management of the socio-economic system. The stages of narrative management and the structure of the business process are presented, as well as the distribution of responsibilities among the participants of the management process. The system of support of narrative management in the socio-economic system is proposed, the architecture of the software package is described. The presented system is implemented. An example of a scenario for managing a socio-economic system using the developed software package is given. The process of managing urban passenger transport is considered as an example. Thus, the study has implemented an approach that allows preserving the observability and manageability of the system and monitoring the process of managing the system, taking into account subject-object interaction. At the same time, the choice of actions to organize management and responsibility for the actions of the participants in the process lies with the participants themselves. The results of actions in the system that may lead to loss of manageability and impossibility of achieving the general goal (destruction of the system) are compensated by modifying the manage strategy without losing the structural coherence of the goal space. This ensures the very existence of the system in its environment.

Keywords: socio-economic system, management support system, management in socio-economic systems, goals model, dynamic goal setting, goals space, system manageability

Data mining methods were used to analyze publication activity of the university staff on the example of the Petrozavodsk State University (hereinafter referred to as PetrSU). In order to identify employees groups with similar indicators of scientific activity, they were clustered. As a result, teaching staff was divided into eight clusters, three of which included employees representing both present and future of science at Petrozavodsk State University, and others that would strive to get into these groups. The presented results of indicators’ statistical processing can be useful for university self-analysis. The university management could draw a conclusion on a current state of scientific activity, both of an individual employee and of the organization as a whole. This will allow to make scientifically-based management decisions in order to improve scientific performance of the organization.

Keywords: university permormance, scientific activity, data mining, clustering, scientometric indicators, h-index, RSCI