This repository contains code for the project my ongoing project exploring the usefulness of modern, `deep', neural networks for forecasting political instability. I still need to do more work related to interpretability (e.g. variable importance and partial dependence plots) and explore different hyper parameter settings for networks with more than 1 hidden layer.
R/: Contains code for different models and plots
python/: Contains code for different neural network models, estimated in Keras
data/: Contains dataset from Ward et al. (2015) article on irregular leadership changes
plots/: Contains figures
poster/: Contains poster presented at the 2016 Annual Meeting of the Society for Political Methodology
