The Export Predictor is a machine learning tool to identify viable export industries and latent comparative advantage. It was designed primarily to help governments in developing countries prioritise their industrial policy. The accompanying website allows users to visualise the predictions, as well as the country proximity networks used in the model.

This repository contains Python scripts to download international trade data via the UN Comtrade API, update a PostgreSQL database, extract features, train and validate the predictive model. The results are visualised using D3 and D3plus. Read more.

The Export Predictor was built by Peter Richens as a project for the General Assembly Data Science course.