This repository contains a model for prediction of civil conflict in fragile nation states across the world. This early warning system for conflict prediction can help governments of such delicate contexts effectively deploy resources so as to save human lives and minimize costs to their economy.
I showcase my analysis considering the case of Pakistan, on account of the country's history of political turmoil in the last decade. My primary source of data is the Armed Conflict Location & Event Dataset, which is a publicly available data repository cataloguing global conflict events on a daily basis.
I have used a Recurrant Neural Network with Long Short-Term Memory for prediction.
Files 1 - 3 explore the data using some visualizations.
File 4 combines multiple input datasets obtained from various sources.
File 5 trains the predictive model (RNN for time series forcasting), tests it, and makes predictions.
File 6 plots the final predictions.