Conflict Forecast

Violent events not only cause immediate suffering but have secondary impacts by preventing access to critical infrastructure and services. 

Disruption could be due to: 

• Physical barriers: destruction of critical infrastructure and services centres. 
• Operational barriers: blockage of infrastructure and shutdown of essential services centres. 

This can lead to access disruption - interruption or restriction of people's ability to reach and use critical infrastructure and services. It also makes it difficult to accurately design operational response and coordinate humanitarian support.

This innovation helps decision-makers visualize conflict events by integrating multiple conflict datasets with socio-economic, population density, critical infrastructure and services data at the subnational level. Using a methodology that combines machine learning (panel forecasting) and natural language processing, Conflict Forecast predicts armed conflict and disruptive events like protests and riots. It correlates conflict predictions with other data layers to understand how risks of conflict, protests and riots relate to access disruptions. The web platform then presents this data in multi-layered maps for resource planning and allocation.

The Conflict Forecast team produces sub-national forecasts of conflict intensity at a 55x55 km resolution. It uses news text from ~6 million articles as a predicting feature and its conflict prediction model performs well. Globally it achieves an accuracy score of 0.97 (1 is perfect accuracy).