Previsão de Insolvência: Uma Estratégia para Balanceamento da Base de Dados Utilizando Variáveis Contábeis de Empresas Brasileiras Bankruptcy Prediction: A Methodology for Balancing Database Using Accounting Variables in Brazilian Companies

The theme of failure prediction is increasingly becoming the subject of new studies and research because it allows for the prediction of a difficult financial situation and also allows for timely measures to reverse this situation by preventing the generation of large social and financial costs. This study also has gained more importance due to changes in business environments, the increase of databases and development of new technologies in computing systems. In Brazil, the studies in this area are still suffering the effect of finding databases of size due to reduced quality of data available, fortunately this situation is changing. Using data coming from the financial statements of Brazilian companies listed on BOVESPA, we present a data mining methodology that addresses the problem of the imbalance of classes, problem exists on this issue because in normal economic environments the number of companies classified as solvents are much larger than those classified as insolvent. This methodology aims to better characterize those companies that have higher potential that they will become insolvent. According to the results of the methodology was successful and could be considered competitive with other methods presented in the literature.