ATTENTION/WARNING - NE PAS DÉPOSER ICI/DO NOT SUBMIT HERE

Ceci est la version de TEST de DIAL.mem. Veuillez ne pas soumettre votre mémoire sur ce site mais bien à l'URL suivante: 'https://thesis.dial.uclouvain.be'.
This is the TEST version of DIAL.mem. Please use the following URL to submit your master thesis: 'https://thesis.dial.uclouvain.be'.
 

Predicting Recovery Rates of Defaulted Bonds Using a Neural Network

Charles de la Brousse, Aymeric
(2018)

Files

charles_de_la_brouse_78521200_2018.pdf
  • Open access
  • Adobe PDF
  • 1.16 MB
Download

Details

Supervisors
Vrins, Frédéric ; Gambetti, Paolo
Faculty
Louvain School of Management
Degree label
Master [120] en Ingénieur de gestion
Abstract
In this study we implement a neural network to forecast recovery rates of defaulted bonds using bond-specific, firm-specific and macroeconomic variables. We benchmark the model with a simple linear regression and a random forest. The network is composed of one hidden layer and a sigmoid activation function to guarantee a prediction in the unit interval. The relative importance of the network's inputs is computed using Garson's and Olden's algorithm. The performance of each model is assessed with and without overestimation penalties. The results suggest that neural networks do not have better predictive abilities given the data and the parameters. Finally, we compare the impact of each model on the Value-at-Risk by simulating 20,000 portfolios. The results show the linear regression computes the highest VaR, followed by the neural network and the random forest.