A data mining approach to the prediction of corporate failure

FY Lin, SI McClean

Research output: Contribution to journalArticle

90 Citations (Scopus)

Abstract

This paper uses a data mining approach to the prediction of corporate failure. Initially, we use four single classifiers - discriminant analysis, logistic regression, neural networks and C5.0 - each based on two feature selection methods for predicting corporate failure. Of the two feature selection methods - human judgement based on financial theory and ANOVA statistical method - we found the ANOVA method performs better than the human judgement method in all classifiers except discriminant analysis. Among the individual classifiers, decision trees and neural networks were found to provide better results. Finally, a hybrid method that combines the best features of several classification models is developed to increase the prediction performance. The empirical tests show that such a hybrid method produces higher prediction accuracy than individual classifiers.
LanguageEnglish
Pages189-195
JournalKnowledge-Based Systems
Volume14
Issue number3-4
DOIs
Publication statusPublished - 23 Jun 2001

Fingerprint

Data mining
Classifiers
Discriminant analysis
Analysis of variance (ANOVA)
Feature extraction
Neural networks
Decision trees
Logistics
Statistical methods
Corporate failure
Classifier
Prediction
Analysis of variance
Hybrid method
Feature selection

Cite this

Lin, FY ; McClean, SI. / A data mining approach to the prediction of corporate failure. 2001 ; Vol. 14, No. 3-4. pp. 189-195.
@article{2f599eac4c2d494093ea1b955f52133c,
title = "A data mining approach to the prediction of corporate failure",
abstract = "This paper uses a data mining approach to the prediction of corporate failure. Initially, we use four single classifiers - discriminant analysis, logistic regression, neural networks and C5.0 - each based on two feature selection methods for predicting corporate failure. Of the two feature selection methods - human judgement based on financial theory and ANOVA statistical method - we found the ANOVA method performs better than the human judgement method in all classifiers except discriminant analysis. Among the individual classifiers, decision trees and neural networks were found to provide better results. Finally, a hybrid method that combines the best features of several classification models is developed to increase the prediction performance. The empirical tests show that such a hybrid method produces higher prediction accuracy than individual classifiers.",
author = "FY Lin and SI McClean",
note = "Other Details ------------------------------------ This paper describes and evaluates a data mining approach to the prediction of corporate failure. It was published, in an earlier version, at the 20th BCS SGES International Conference in December 2000, where it won a best paper award and was selected to appear as a journal paper in JKBS. The paper combines classification algorithms with business knowledge and human judgment. The results were evaluated, and shown to give good results, using a standard online database on corporate accounts. The topic is part of ongoing work of the Information and Software Engineering research group in the area of knowledge discovery.",
year = "2001",
month = "6",
day = "23",
doi = "10.1016/S0950-7051(01)00096-X",
language = "English",
volume = "14",
pages = "189--195",
number = "3-4",

}

A data mining approach to the prediction of corporate failure. / Lin, FY; McClean, SI.

Vol. 14, No. 3-4, 23.06.2001, p. 189-195.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A data mining approach to the prediction of corporate failure

AU - Lin, FY

AU - McClean, SI

N1 - Other Details ------------------------------------ This paper describes and evaluates a data mining approach to the prediction of corporate failure. It was published, in an earlier version, at the 20th BCS SGES International Conference in December 2000, where it won a best paper award and was selected to appear as a journal paper in JKBS. The paper combines classification algorithms with business knowledge and human judgment. The results were evaluated, and shown to give good results, using a standard online database on corporate accounts. The topic is part of ongoing work of the Information and Software Engineering research group in the area of knowledge discovery.

PY - 2001/6/23

Y1 - 2001/6/23

N2 - This paper uses a data mining approach to the prediction of corporate failure. Initially, we use four single classifiers - discriminant analysis, logistic regression, neural networks and C5.0 - each based on two feature selection methods for predicting corporate failure. Of the two feature selection methods - human judgement based on financial theory and ANOVA statistical method - we found the ANOVA method performs better than the human judgement method in all classifiers except discriminant analysis. Among the individual classifiers, decision trees and neural networks were found to provide better results. Finally, a hybrid method that combines the best features of several classification models is developed to increase the prediction performance. The empirical tests show that such a hybrid method produces higher prediction accuracy than individual classifiers.

AB - This paper uses a data mining approach to the prediction of corporate failure. Initially, we use four single classifiers - discriminant analysis, logistic regression, neural networks and C5.0 - each based on two feature selection methods for predicting corporate failure. Of the two feature selection methods - human judgement based on financial theory and ANOVA statistical method - we found the ANOVA method performs better than the human judgement method in all classifiers except discriminant analysis. Among the individual classifiers, decision trees and neural networks were found to provide better results. Finally, a hybrid method that combines the best features of several classification models is developed to increase the prediction performance. The empirical tests show that such a hybrid method produces higher prediction accuracy than individual classifiers.

U2 - 10.1016/S0950-7051(01)00096-X

DO - 10.1016/S0950-7051(01)00096-X

M3 - Article

VL - 14

SP - 189

EP - 195

IS - 3-4

ER -