The relationships between armed groups in North and Western Africa are often characterized by widely varying rivalries and alliances. Groups that were fighting each other one day can very well be allied the next day.
In a recent paper published in Terrorism and Political Violence, my co-authors Christian Leuprecht and David Skillicorn and I have used network analysis to represent alliances and conflicts among 179 organizations involved in violence in the region. Building on ACLED data from 1997-2014, we show that the structural positions of armed groups affect their ability to resort to political violence. In other words, groups with similar allies and foes have similar patterns of violence.
Our new paper combines, for the first time, two spectral embedding techniques that have previously been considered separately: one that takes into account the direction of relationships between belligerents, and one that takes into consideration whether relationships between groups are positive or negative.