Onderwijstype
Machine Learning for socio-economic data using Python - 4 ECTS
Registration
Content
A new methodological wave in the form of Machine Learning tools currently freshens up empirical research in social science. Classic statistical and econometric toolboxes are rapidly complemented with new data-driven algorithms that search for patterns in often large datasets to yield better forecasting or classification models (Storm et al., 2020; Brignoli et al., 2024). This course discusses some of these new ML techniques and how they can be used to analyse socio-economic data that has an explicit time component, e.g. time-series data such as prices or macro-economic data, and panel data collected from households or firms over multiple years. How do these new methods deal with the time dimension of the data? How do they differ from classic econometric methods for time series (e.g. ARIMA, VAR, VEC models) or panel data (e.g. mixed effect models). The ML methods discussed are LASSO, random trees and forests, and neural networks.
Every day a topic is first introduced in a two-hour lecture in the morning. Every afternoon participants work on an exercise in which they use Python to code the respective method using a given dataset. Where useful, the ML coded problems are compared with classic econometric coding of the respective problem in Python.
Learning outcomes
After successful completion participants are expected to be able to:
- explain how Machine Learning techniques can be useful for econometric tasks;
- compare Machine Learning data analysis with classic time series and panel data analysis
- program econometric and machine learning time-series and panel data models in Python
- evaluate Machine Learning time series and panel data analysis
Target group and min/max number of participants
PhD candidates from AEP, BEC, BMO, DEC, ENR, MCB, RHI, UEC plus interested MME-D students.
Minimum: 8 – Maximum: 15
Assumed prior knowledge
AEP33806 Advanced Econometrics or equivalent
Assessment
Short paper and Python code describing the analysis of an own time series or panel data problem.
Schedule
| Session 1 | 2 March |
| Session 2 | 3 March |
| Session 3 | 4 March |
| Session 4 | 5 March |
| Session 5 | 6 March |
Lecture time in the mornings are 9.30-11.15. Afternoon lab sessions are 13.30-16.00.
Course fees
| WGS PhD's with TSP | 300 euro |
| a) All other PhD candidates b) Postdocs and staff of the above mentioned Graduate Schools | 640 euro |
| all others | 900 euro |
Cancellation conditions
Participants can cancel their registration free of charge 1 month before the course starts. A cancellation fee of 100% applies if a participant cancels his/her registration less than 1 month prior to the start of the course.
The organisers have the right to cancel the course no later than one month before the planned course start date in the case that the number of registrations does not reach the minimum.
The participants will be notified of any changes at their e-mail addresses.