Aplicação de redes neurais na previsão da captação líquida do mercado de previdência aberta brasileiro

• Autor(a) principal: Pinto, Arizoly Rodrigues
• Data de Publicação: 2009
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional do FGV (FGV Repositório Digital)
• Texto Completo: http://hdl.handle.net/10438/4300
• topic: Finanças; Redes neurais; Investimentos; Previdência aberta; Processo decisório; Captação líquida; Economia; Redes neurais (Computação); Investimentos - Processo decisório; Previdência privada - Brasil

The objective of this work is to explore the use of Neural Networks in the Prediction of the Brazilian Net Inflow of Private Pension Fund Market, as a tool for decision-making and support to the management of companies. To build this model was used neural networks, a tool that is proving suitable for use in nonlinear models with better results than other techniques. The main data source for this work was the FENAPREVI - National Federation of Private Pension and Life. For comparison with the model of neural networks, we used a model of Multiple Linear Regression as a benchmark, in order to demonstrate the suitability of the tool in view of the objectives outlined in this job. The model was constructed from the monthly industry information, between May 2002 and August 2009, considering what has been called 'the live market”, which include products PGBL and VGBL, marketed without interruption during this period for the EAPP – Entities Open Private Pension Funds. The results demonstrated the suitability of the tool Neural Networks, which obtained better results than those obtained using Multiple Linear Regression