Forecasting Country Stability in North Africa

We develop a novel approach to predict certain type of stability events (battles, battles won by a government, riots/protests, violence against civilians) in countries by monitoring the content of a mix of traditional news, blog, and social media data. Specifically, we show that by monitoring sentiment on both pro- and anti-government entities within a country, even with a relative paucity of longitudinal data (36 time points), we can predict these stability related events with just over 80% classification accuracy. We report on our methods, together with a description of a prototype system called Sentibility that tracks country stability related events. In addition, we cast light on the key entities, sentiments on whom were correlated strongly (positively or negatively) by both Pearson and Spearman correlation coefficients, with such stability events in 3 countries: Egypt, Morocco, and Sudan.