The research is aimed at building a methodology detection of one of the new and extremely destructive social mechanisms for mobilizing political protest in the information and network society, which is the mechanism for mediating local incidents. Based on algorithms artificial intelligence it is planned to develop a methodology and software for diagnosing local incidents in social network. If this methodology is used, it will be possible to study a new and extremely dangerous phenomenon, which is a mass mobilization of protest caused not by endogenous, but exogenous factors, in some cases inspired by external stakeholders, more strictly and correctly. A distinctive feature of the proposed approach is the possibility of developing an automated system based on specially developed algorithms for monitoring potentially dangerous local incidents for socio-political stability. In the work is received information about the functions and potential of various machine learning models. Is made conclusion about which models are the most optimal for analyzing messages about protest activity. The methodology proposed by the authors is promising for further research and development, because it makes it possible to convert text values into numeric values and to add a sign of the presence of a call to messages.