PeaceKeeping

Skills & Tools

• R

• Python

• HSMM

• XGBoost

• Shiny

Predicting the likelihood of events in Mali

In a conflict, knowing when and where the next event will take place can help prevent humanitarian hardship. Our goal was to predict when and in which region of Mali a given event was likely to occur.

We used GDELT and ACLED datasets to gather information from news sources and identify events (as defined by the ACLED dataset) and weather data gathered from the Weather Company. With these data we built XGBoost and Hidden Semi-Markov models to predict the likelihood of an event (Battle, Strategic Developments, ect.) in each administrative division within a given timeframe. These results were presented in an interactive Shiny dashboard, allowing the user to filter by timeframe and event type.