Developing an intelligent system for predicting bankruptcy

SAMAR ALY; MARCO ALFONSE; MOHAMED I. ROUSHDY; ABDEL-BADEEH M. SALEM

Author(s)

SAMAR ALY Computer Science Department, Faculty of Computer and Information Science, Ain Shams University, Cairo, Egypt

MARCO ALFONSE Computer Science Department, Faculty of Computer and Information Science, Ain Shams University, Cairo, Egypt

MOHAMED I. ROUSHDY Faculty of Computers & Information Technology, Future University in Egypt

ABDEL-BADEEH M. SALEM Computer Science Department, Faculty of Computer and Information Science, Ain Shams University, Cairo, Egypt

Abstract

Bankruptcy prediction is a very important metric for making economic decisions. It is a very critical element in credit risk analysis. Machine learning based classifiers have been heavily utilized in predicting bankruptcy. In terms of machine learning, predicting bankruptcy with imbalanced dataset is a very big challenge. Despite the current existing models for bankruptcy prediction, finding a model that achieves a high-performance measurement with imbalanced datasets is still an interesting point of research. The reason behind this challenge is the fact that imbalanced dataset leads to misclassification results. This paper aims to develop a robust machine learning based model for predicting bankruptcy with solving imbalanced dataset problem. Four different re-sampling strategies were applied to solve the imbalanced class distribution problem based on three popular datasets. The used datasets were downloaded from the University of California, Irvine (UCI) machine learning repository. The developed model based on different single and ensemble machine learning classifiers. The overall experimental results showed that the best performance of the developed model with the Polish dataset was 97% for accuracy and 95.4% for AUC (Area Under the Curve). Moreover, the best performance with the Australian dataset was 88.4% for accuracy and 92% for AUC. And the best performance with the German dataset was 81.5% for accuracy and 83.4% for AUC ratio.

Publication Details

Page(s) 2068-2088

DOI DOI not available

Published Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 7, Year: 2022

Request the full text from PASTIC

Keywords

Imbalanced Dataset, Artificial Intelligence, Machine Learning Techniques, Bankruptcy prediction, Ensemble Techniques