Abstract

REAL-TIME POLITICALLY MOTIVATED VIOLENCE EARLY WARNING
Track: Homeland and National Security
Authors: Nelson Okello

Election related violence in Kenya has recurred in 75% of the general election years. The worst of these was the 2007 general election, in which approximately 1,000 people lost their lives and more than 350,000 displaced after a disputed general election. Security apparatus are often ill equipped for these circumstances due to resource constraints and uncertainty. Using GIS and social network data, we designed a real-time early warning system which feeds security agencies with the information they need to avert potential violence. Using machine learning algorithms, we are able to aggregate potential violence cases by location. Using spatial routing, we search and inform security agencies nearest to the potential violence areas. The system have mobile and web clients. We'll put this system into use in March, 2013, general elections. We'll release this software under an open source license for the good of the wider community.