Aplicação de redes neurais na classificação de rentabilidade futura de empresas

• Autor(a) principal: Matsumoto, Élia Yathie
• Data de Publicação: 2008
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional do FGV (FGV Repositório Digital)
• Texto Completo: http://hdl.handle.net/10438/2628
• topic: Finanças; Classificação de padrões (Computação); Economia; Redes neurais (Computação); Investimentos - Processo decisório; Empresas - Lucratividade

The motivation of this work is to verify the efficiency of neural networks as a tool for classifying forecasts of companies’ return on equity, so that they can be used in order to provide support for the development of investment decision support systems. The results obtained by the proposed neural networks mode are compared to those obtained by the use of the classic multiple linear regression as a minimum reference and, as a benchmark, to those obtained via ordinal logistic multiple regression. In this work, we gathered top 1000 companies’ financial and accounting data, annually listed by Melhores e Maiores – Exame publication (Editora Abril), from 1996 to 2006. The three models were built using data relative to the period between 1996 and 2004. Using the 2005 data as input in order to forecast companies’ classification in 2006, the three models’ outputs were compared to the observed 2006 classification, and the neural network model yielded the best results.