Bankruptcy Prediction Using First-Order Autonomous Learning Multi-Model Classifier

Amine Sabek, Jakub Horák, Hussam Musa, Amélia Ferreira da Silva

Bankruptcy Prediction Using First-Order Autonomous Learning Multi-Model Classifier

Číslo: 4/2024
Periodikum: Statistika
DOI: 10.54694/stat.2024.30

Klíčová slova: Bankruptcy prediction, first-order, autonomous learning, Multi-Model Classifier, Principal Components Analysis

Pro získání musíte mít účet v Citace PRO.

Přečíst po přihlášení

Anotace: Research background: Bankruptcy and financial distress prediction has always been an integral part of any financial management system. It gives an indication to stakeholders to take precautionary measures in order to avoid losses. The traditional approaches for prediction, including logistic regression and discriminant analysis, are constrained by their inability to deal with complex and high-dimensional data (Odom and Sharda, 1990; Min and Lee, 2005). Recent developments in the field of machine learning, and particularly autonomous learning classifiers, present a potential proposed alternative.

Purpose: The purpose of this paper is to propose a first-order autonomous learning classifier (F-O ALMM0) for predicting bankruptcy of business entities and individuals.
Design/methodology/approach: The data file contained a total of 352 companies obtained from the Kaggle database and incorporating 83 financial ratios. Initially, the model's performance was assessed as a preliminary step, but the results were average, followed by the application of Principal Component Analysis (PCA) to enhance the quality of the input’s variables. Afterwards, the number of independent variables was reduced to 26. Thus, the results were improved.