Designing a human-machine interface is an invisible and important task in the software security life cycle. Depending on the results of the project of the deposit, the rhubarb of the dryness of the coristuvachіv pіd hour of the vikorіstanny of the product. At the moment, there are different approaches to evaluating the usability of sites with an inspection of the content of the page, their design, the placement of elements of cura-tion, etc. But a complex software product that evaluates the bi-usability of any pro-ponated site, so far nothing. Thus, the task of automated website usability evaluation is relevant. The method of robots is the development of system interfaces for websites based on fuzzy logic. For reachability, the following tasks were decided: 1) to make lin-guistic changes to assess the usability of the site interface; 2) to develop and implement the algorithm of the parser of the conditional code of the HTML page for constructing the terms of linguistic snakes; 3) compiling a questionnaire for checking usability on the site for experts and providing a questionnaire; 4) development of an algorithm for fuzzy derivation of assessments of the usability of the site interface; 5) development of a data-base of estimates; 6) to develop, implement and test at a strict level programs for evalu-ating usability in the interface of sites. The starting speed of a website without mediation is entered by the number of its guard assistants, for sale, as such, the functioning is transferred to this site, to the sat-isfaction level of assistants from robots from the site. The movement of usability can be carried out according to formal evaluation criteria, as linked to certain system rules of fuzzy logic. As part of the robot, a system was developed that determines usability in the site interface based on fuzzy logic rules.

The work is devoted to the statistical analysis of the text and the study of the dynamics of classification. In the work, the selection of statistical features of the text, the classification of texts belonging to different authors, and the study of the dynamics of classification accuracy depending on the length of text fragments are carried out. To solve the problem, the following methods were used: natural language processing methods; statistical characteristics of texts; machine learning methods; dimensionality reduction methods for visualization capability. On the basis of the obtained dynamics of changes in classification accuracy depending on the lengths of text fragments, appropriate conclusions were drawn regarding the optimal length of texts used for training and testing models. The task was solved in the Jupyter Notebook software environment of the Anaconda distribution, which allows you to immediately install Python and the necessary libraries.