Geography of Terror


This project aims to predict spikes in national terrorist activity two years in advance to allow governments time to develop preventative social policies.

The project uses terror incident data from The Global Terrorism Database (GTD) and climate data from Berkeley Earth.

Time-series data is tackled by feature engineering rolling metrics, and predictions are made by training algorithms XGBoost, Tree Ensemble Gradient Boosted Classifiers and K-Nearest Neighbors.

Find some of the code used in the project on Github and check out the D3 Visualization below!