UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

Supervised machine learning algorithms for the estimation of the probability of default in corporate credit risk

Sariev, Eduard; (2021) Supervised machine learning algorithms for the estimation of the probability of default in corporate credit risk. Doctoral thesis (Ph.D), UCL (University College London).

[thumbnail of Thesis_1037961.pdf] Text
Thesis_1037961.pdf - Accepted version
Access restricted to UCL open access staff until 1 March 2022.

Download (2MB)

Abstract

This thesis investigates the application of non-linear supervised machine learning algorithms for estimating Probability of Default (PD) of corporate clients. To achieve this, the thesis is separated into three different experiments: 1. The first experiment investigates a wrapper feature selection method and its application on the support vector machines (SVMs) and logistic regression (LR). The logistic regression model is the most popular approach used for estimating PD in a rich default portfolio. However, other alternatives to PD estimation are available. SVMs method is compared to the logistic regression model using the proposed feature selection method. 2. The second experiment investigates the application of artificial neural networks (ANNs) for estimating PD of corporate clients. In particular ANNs are regularized and trained both with classical and Bayesian approach. Furthermore, different network architectures are explored and specifically the Bayesian estimation and regularization is compared to the classical estimation and regularization. 3. The third experiment investigates the k-Nearest Neighbours algorithm (KNNs). This algorithm is trained using both Bayesian and classical methods. KNNs could be efficiently applied to estimating PD. In addition, other supervised machine learning algorithms such as Decision trees (DTs), Linear discriminant analysis (LDA) and Naive Bayes (NB) were applied and their performance summarized and compared to that of the SVMs, ANNs, KNNs and logistic regression. The contribution of this thesis to science is to provide efficient and at the same time applicable methods for estimating PD of corporate clients. This thesis contributes to the existing literature in a number of ways. 1. First, this research proposes an innovative feature selection method for SVMs. 2. Second, this research proposes an innovative Bayesian estimation methods to regularize ANNs. 3. Third, this research proposes an innovative Bayesian approaches to the estimation of KNNs. Nonetheless, the objective of the research is to promote the use of the Bayesian non-linear supervised machine learning methods that are currently not heavily applied in the industry for PD estimation of corporate clients.

Type: Thesis (Doctoral)
Qualification: Ph.D
Title: Supervised machine learning algorithms for the estimation of the probability of default in corporate credit risk
Event: UCL
Language: English
Additional information: Copyright © The Author 2021. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request.
UCL classification: UCL
UCL > Provost and Vice Provost Offices
UCL > Provost and Vice Provost Offices > UCL BEAMS
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science
URI: https://discovery.ucl.ac.uk/id/eprint/10122087
Downloads since deposit
2Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item