Predicting Social Unrest Using GDELT

Abstract

Social unrest is a negative consequence of certain events and social factors that cause widespread dissatisfaction in society. We wanted to use the power of machine learning (Random Forests, Boosting, and Neural Networks) to try to explain and predict when huge social unrest events (Huge social unrest events are major social unrest events as recognized by Wikipedia page ‘List of incidents of civil unrest in the United States’) might unfold. We examined and found that the volume of news articles published with a negative sentiment grew after one such event - the death of Sandra Bland - and in other similar incidents where major civil unrest followed. We used news articles captured from Google’s GDELT (Global Database of Events, Language, and Tone) table at various timestamps as a medium to study the factors and events in society that lead to large scale unrest at both State and County levels in the United States of America. In being able to identify and predict social unrest at the county level, programs/applications can be deployed to counteract its adverse effects. This paper attempts to address this task of identifying, understanding, and predicting when social unrest might occur.