Aplicaciones en Economía del Aprendizaje AutomáticoMachine Learning Applications in Economics

  1. Matthew Smith
Supervised by:
  1. Francisco Álvarez González Director

Defence university: Universidad Complutense de Madrid

Fecha de defensa: 06 May 2022

  1. Miguel Jerez Méndez Chair
  2. Alfredo Garcia-Hiernaux Secretary
  3. Pablo del Río González Committee member
  4. Stefano Sacchetto Committee member
  5. Katja Gilly de la Sierra-Llamazares Committee member

Type: Thesis


This Thesis examines problems in economics from a Machine Learning perspective. Emphasisis given on the interpretability of Machine Learning algorithms as opposed to blackbox predictions models. Chapter 1 provides an overview of the terminology and Machine Learning methods used throughout this Thesis. This chapter aims to build a roadmap from simple decision tree models to more advanced ensemble boosted algorithms. Other Machine Learning models are also explained. A discussion of the advances in Machine Learning in economics is also provided along with some of the pitfalls that Machine Learning faces. Moreover, an example of how Shapley values from coalition game theory are used to help infer inference from the Machine Learning models' predictions. Chapter 2 analyses the problem of bankruptcy prediction in the Spanish economy and how Machine Learning, not only provides more predictive accuracy, but can also provide adierent interpretation of the results that traditional econometric models cannot. Several financial ratios are constructed and passed to a series of Machine Learning algorithms. Case studies are provided which may aid in better decision-making from financial institutions. A section containing supplementary material based on further analysis is also provided...