Using GDELT to forecast violence in Afghanistan

The Global Dataset of Events, Location, and Tone (GDELT) is a new, 230 million (and growing daily) is the first ever machine-coded political event data dataset to provide information on event location.  For those attending ISA, Kalev Leetaru and Phil Schrodt will be formally introducing the GDELT dataset.  The full dataset will be publicly available soon, but for now you can access an older version here.

From a forecasting perspective, the benefits of a machine-coded dataset updated in (near) real-time that provides specific latitude-longitude coordinates are numerous.  In the first ever empirical analysis using GDELT (pdf of paper –> “Predicting Future Levels of Violence in Afghanistan Districts with GDELT“), I build an empirical model that predicts the level of conflict at the district-month level in Afghanistan.  Below is .gif that Joshua Stevens built using GDELT that reflects the distribution of conflict events in Afghanistan over time.

Comments are closed.