Online Training Workshop on Machine Learning for Economic Analysis

20-21 and 27-28 May 2024

Objectives: The objective of the workshop is to train participants in the main Machine Learning (ML) methods in the specific context of economic analysis. The training will focus on supervised and semi-supervised classification techniques. In this context, the aim will be to propose a reflection on the existence of a trade-off between the interpretability and predictive performance of models derived from ML algorithms. This intermediate-level course is aimed at participants with notions of statistical learning and basic notions of programming. 

Knowledge transfer: the course will combine a general presentation of ML with a more detailed presentation of a few methods commonly used in the context of economic analysis.

Public: This course is aimed at professionals and students who wish to use ML methods in their economic analysis. The applications will be based on Python software using Google Colab (no installation is required before the workshop).

The workshop will be conducted mainly in English, with the possibility of using French if needed. The workshop will be held over Zoom. The training will be conducted by Prof. Christophe Hurlin.

Eligibility:

Applicants should be:

• Researchers from the ERF region with MAs or PhDs or in the process of completing these degrees.
• Less than 40 years of age.
• Trained in statistics, economics and mathematics and able to use statistical programs.
• Familiar with the management and analysis of macroeconomic datasets.
• Priority will be given to those who have not attended previous ERF workshops. 

Application procedure:

Applications must include a curriculum vitae (CV) with a minimum of one reference name and a motivation letter indicating why the candidate is interested in the topic of the workshop and how it relates to his/her research.

Deadline for submissions: April 30, 2024.

Send your submission through this link

Program:

Day 1 (May 20)

16:00-16:30 – Introduction and Objectives of the Workshop
16:30-17:30 – General introduction to ML

• ML and artificial intelligence.
• The different types of machine learning.
• The main ML algorithms.
• ML algorithm construction process.
• Bias/variance trade-off

17:30-17:45 – Coffee Break
17:45-19:00 – Regression and classification trees

• Introduction to regression and classification tree.
• Building classification trees: CART algorithm.
• Building regression trees.
• Applications of classification and regression trees.
• Post and pre-pruning methods.
   

Day 2 (May 21)

16:00-17:30 – General introduction to ML (part 2)

• Performance evaluation of ML algorithms: evaluation criteria and methodology.
• Trade-off between complexity and performance.
• Choosing hyperparameters by cross-validation: controlling the risk of over-fitting.
• The issue of unbalanced samples. 

17:30-17:45 – Coffee Break
17:45-19:00 – Supervised learning methods: Bagging

• Principle of Bagging methods.
• Advantages and limitations of Bagging.
• The case of random forest.
   

Day 3 (May 27)

16:00-17:30 – Supervised learning methods: Boosting

• Principles of boosting methods.
• Advantages and limitations of boosting.
• AdaBoost method.
• Gradient Boosting and XGBoost methods.

17:30-17:45 – Coffee Break
17:45-19:00 – Supervised learning methods: SVM

• Introduction to the support vector machine (SVM).
• Formalization of the SVM.
• The kernel trick.
• Support Vector Regression (SVR).
   

Day 4 (May 28)

16:00-17:30 – Semi-supervised learning methods

• Principle of semi-supervised learning
• Pseudo labelling
• Self-training methods

17:30-17:45 – Coffee Break
17:45-19:00 – Applications

• Credit scoring model
• Forecasting housing prices