Introduction to Machine Learning for Economists

This repository contains the teaching material of the machine learning course I am teaching on the master level at the University of Bordeaux. This course has a strong a focus on the utilization of machine learning for impact evaluation and a first approach to remote sensing.

Course description

This course aims to cover the fundamental concepts of machine learning in economics and their practical applications. It aims to provide an introduction to basic and key principles of machine learning, as well as potential utilization for economists. This course will not cover any details of algorithms and computational issues – there are many specialised courses from Computer Science that go into these details.

You can find the syllabus here.

Assignment

There will be a problem set due at the end of the semester. More details to come.

Main references

I am relying heavily on the James, Witten, Hastie, Tibshirani, and Taylor (2023)'s textbook for explaining the key concepts and for the first lab sessions. You can find the textbook directly on their website.

In addition, I relied on Matteo Courthoud for double selection and DDML, Paul Schrimpf's notes to implement General Machine Learning, and the Mixtape's codes for causal forests.

Lectures

1 Class overview and technical assistance

• Check my Github page to set up python on your device. Then follow the steps I mention here to open start your environment.
• Lab: quick intro to python

2 Introduction to Machine Learning

• Statistical learning, model fit and bias-variance trade-off. Lab
• Cross-validation. Lab
• Regularized regression Lab and tree-based method Lab

3 Average treatment effects with too many control variables

• Double selection and Double Debiased Machine Learning Lab

4 Estimating heterogeneous treatment effect

• Causal Trees/Forests and Generic Machine Learning Lab

5 Other applications: introduction to spatial data Lab