Abstract
The fight against extremism in social networks on the internet can be classified as one of the most important, mandatory and difficult tasks, but one that has a solution. The development of effective systems and models for the detection and Prevention of political extremism is the basis for ensuring security and stability in the country. Accordingly, it requires conducting scientific research in this area, thereby implementing an effective and timely set of measures aimed at identifying, preventing and suppressing any manifestations of extremism. A number of machine learning models have been studied. Measures for their implementation and careful evaluation were carried out. Machine learning models allow a comprehensive comparison of their performance in this area. Conducting this analysis contributed to the identification of the most effective methods, achieving an optimal balance between accuracy, efficiency and practicality. Machine learning models a comparative analysis of the methods of LSTM, TF-IDF and KNN showed their features and limitations in determining extremism in textual content. LSTM has shown high performance with consistent data, while TF-IDF has shown good performance in producing features, while KNN has proven effective for simple classification tasks. This assessment gave a clear idea of their suitability for various aspects of the problem and helped to choose the most optimal approach.