Methods of machine learning in determining mental state based on data from a coordinate input device

The subject of the study is the potential use of machine learning methods to analyze computer mouse movements for diagnosing users' mental states. The proposed model, based on data about mouse movement speed, click frequency, and aiming accuracy, allows for the assessment of users' anxiety and stress levels. The experiment is conducted within the framework of the Remote Topology project, which collects data for subsequent model validation. The main focus is on automating the monitoring process of learners' psycho-emotional states and utilizing modern technologies to support their psychological well-being in educational environments. The developed approach involves performing tasks within a comprehensive infrastructure, enabling participants to work with standard system administrator tasks. The typical assignments incorporate logic whereby the task will not be solved using conventional algorithms. The research method is the implementation of machine learning techniques in the psychodiagnostics of students, allowing for the analysis of big data, the identification of hidden patterns, and the forecasting of emotional states. Examples of projects and systems utilizing NLP, movement analysis, and activity for the early detection of anxiety, stress, and depression are provided. The main outcome of the conducted research is the statistical analysis of stress and anxiety among participants. The experimental protocol was designed to recreate a scenario similar to the workday of a mid-level enterprise specialist solving tasks according to the professional standard 6.027, a specialist in the administration of network devices of information and communication systems. As a result, the experimental protocol induces an overall increase in perceived stress and anxiety levels, which can be accurately determined using machine learning methods. The differences in the tasks of these classes and algorithms complicate direct comparisons of the results obtained in our study with existing studies presented in the literature review. During previous and current research aimed at determining stress levels using mouse activity data, validation of the experimental group was conducted using machine learning methods. The analysis confirmed the significance of such features as key hold time, intervals between key presses, and mouse movement speed, which aligns with the findings of previous works.