This study focuses on the comparison and selection of the similarity coefficient for comparing social circles of users. Comparison of the social environment is used in comparing profiles in various social networking cites and identifying those that belong to one user. Six similarity coefficients are compared using machine learning methods and mathematical statistics. The selected coefficient will allow us to more accurately compare social circles of users and increase efficiency of the binary classifier (logistic regression) determining pair of profiles in various social networking cite belonging to one user. In addition, results of this study can be used in the analysis of social networking cites, for example, in the tasks of social computing.