Recurrent Neural Networks for volatility estimation - a comparative study of machine learning and traditional methods for volatility estimation

Detta är en C-uppsats från Handelshögskolan i Stockholm/Institutionen för finansiell ekonomi

Sammanfattning: Financial decisions are largely based on a tradeoff between risk and return. While the definition of risk is not equal to volatility, it is often used as a proxy for it. Hence, volatility forecasting is of great importance and an essential part of asset pricing, portfolio optimization, and risk management. The purpose of this paper is to investigate if a Recurrent Neural Network could provide more precise estimations of seasonal volatility and if so, how it compares to other commonly used models. We prove that they do provide good estimations as well as outperform the other models in doing so.

  HÄR KAN DU HÄMTA UPPSATSEN I FULLTEXT. (följ länken till nästa sida)