Feature selection for bankruptcy prediction : a multi-objective optimization ap...

• Autor(es): Mendes, F. ; Duarte, J. ; Vieira, Armando ; Gaspar-Cunha, A. ; Ribeiro, Bernardete ; Ribeiro, André M. S. ; Neves, João Carvalho
• Data: 2010
• Identificador Persistente: http://hdl.handle.net/1822/19288
• Origem: RepositóriUM - Universidade do Minho
• Assunto(s): Feature selection; Bankruptcy prediction; Multi-objective optimization; Evolutionary algorithms; Support vector machines; Logistic regression
• Tipo de Documento: Documento de conferência
• Idioma: Inglês

Descrição

In this work a Multi-Objective Evolutionary Algorithm (MOEA) was applied for feature selection in the problem of bankruptcy prediction. The aim is to maximize the accuracy of the classifier while keeping the number of features low. A two-objective problem - minimization of the number of features and accuracy maximization – is fully analyzed using two classifiers: Support Vector Machines and Logistic Function. A database containing financial statements of 1200 medium sized private French companies was used. It was shown that MOEA is a very efficient feature selection approach. Furthermore, it can provide very useful information for the decision maker in characterizing the financial health of a company.