The paper addresses the problem of automated detection of destructive verbal impacts in user-generated content of digital platforms as an element of information security assurance. A method for context-semantic identification of aggressive and discriminatory statements based on the RuBERT transformer model fine-tuned on a specialized annotated corpus of Russian-language messages is proposed. The procedures of data preparation, training of a binary classifier, and probabilistic interpretation of the results are described. Experimental evaluation confirms the effectiveness and robustness of the method with respect to lexical variability and context-dependent forms of verbal aggression, as well as the possibility of its integration into automated systems for monitoring and protection of the information space.

Keywords: information security, destructive content, verbal aggression, automatic moderation, context-semantic analysis, transformer model, RuBERT, binary classification, machine learning, natural language processing, monitoring system, intelligent filtering

The article focuses on the application of machine learning methods for predicting failures in industrial equipment. A review of modern approaches such as Random Forest, SVM, and XGBoost is presented, with emphasis on their accuracy, robustness, and suitability for engineering tasks. Based on the analysis of real-world data (temperature, pressure, vibration, humidity), models were trained and compared, with XGBoost demonstrating the best performance. Key parameters influencing failures were identified, and a recommendation system was proposed, combining statistical analysis and predictive modeling. The developed solution enables timely detection of failure risks and optimization of maintenance processes.

Keywords: machine learning, predictive modeling, equipment management, failure prediction, data analysis