In research for NATO’s StratCOM, we studied the market for social-media manipulation using metrics such as the selling of likes, comments, shares, accounts, and followers. Using cluster analysis, we studied a host of factors and defined three categories of social media manipulation that became subjects of further research.
In this case, clustering was used to recover missing data in a database and to fill in typical cluster values.
In research into the effects of banning VKontakte in Ukraine, we used clustering in our analysis of post topics. This helped us conclude that ‘Pro-Russian propaganda’ notably increased while the share of ‘Ukrainian news’ decreased.
In NATO StratCom research on the virtual Russian world of the Baltics, we analyzed more than 92 million posts and millions of user profiles. Based on a cluster analysis of user profiles, it was possible to identify four types of ideological users: Writers, Distributors, Readers, and Members of the Active Reserve. As a result, we managed to study them more accurately.
In research conducted with Internews Ukraine on the Ukrainian presidential elections, our script determined 152 topic clusters in more than 9K texts. Thanks to clustering analysis, we identified 40 users who exhibited strange activity on VKontakte and appeared to be social bots.
In a comparative study of two Cesarean techniques, the same methods can’t be applied to different types of patients. In this study, clusters of patients were identified at the beginning so that further research on homogeneous groups could be conducted.