Login | Register

Bankruptcy Prediction by Deep Learning and Machine Learning Methods


Bankruptcy Prediction by Deep Learning and Machine Learning Methods

Zahiri, Parisa (2022) Bankruptcy Prediction by Deep Learning and Machine Learning Methods. Masters thesis, Concordia University.

[thumbnail of Zahiri_MASc_F2022.pdf]
Text (application/pdf)
Zahiri_MASc_F2022.pdf - Accepted Version
Available under License Spectrum Terms of Access.


Bankruptcy prediction plays a crucial role in today’s businesses to survive in a competitive world. For avoiding the risk of bankruptcy, researchers have conducted significant research in field of artificial intelligence for predicting bankruptcy. However, the performance of deep learning methods is not well understood. To address this research gap, we make the following main contributions: We applied deep learning methods into Polish datasets in addition to traditional machine learning techniques. We applied several versions of convolutional neural networks and artificial neural networks to several datasets created from the available dataset. Specifically, we created 5 extra datasets for each year in addition to the entire datasets for five years. We incorporated some techniques to balance the datasets and measured the impacts these techniques have on performance measures. This step is important because the datasets are imbalanced, i.e., the proportion of firms experiencing bankruptcy is much lower than those who did not go bankrupt. For deep learning techniques, we also explored preprocessing approaches and measured their impacts on results. Specifically, we used validation on the same datasets of studies in the literature and compared our results with those available in the literature with the same test bed. Our results shed light on the impact of preprocessing and balancing techniques in deep learning, as well as different architectures for deep learning methods. We observed improvement, compared to the literature, in terms of accuracy and provided insights on the value of different deep learning architectures and preprocessing on the sensitivity of the results.

Divisions:Concordia University > Gina Cody School of Engineering and Computer Science > Mechanical, Industrial and Aerospace Engineering
Item Type:Thesis (Masters)
Authors:Zahiri, Parisa
Institution:Concordia University
Degree Name:M.A. Sc.
Program:Industrial Engineering
Date:1 November 2022
Thesis Supervisor(s):Kazemi Zanjani, Masoumeh and Lahmiri, Salim
ID Code:991305
Deposited By: Parisa Zahiri
Deposited On:21 Jun 2023 14:41
Last Modified:21 Jun 2023 14:41
All items in Spectrum are protected by copyright, with all rights reserved. The use of items is governed by Spectrum's terms of access.

Repository Staff Only: item control page

Downloads per month over past year

Research related to the current document (at the CORE website)
- Research related to the current document (at the CORE website)
Back to top Back to top