Décima Semana Integrada do Instituto de Física de São Carlos

Using neural networks to forecast stock prices using simulated data

Não agendado

20m

Permito a publicação do pôster e vídeo Termodinâmica, Fluidodinâmica e Estatística Apresentação de Pôsteres

Descrição

This project is focused on using neural networks architectures in the forecast of stock prices, with it's main contribution being the creation of a simulated dataset, aimed in investigating it’s effectiveness when applied in real stock prices of assets located in BOVESPA. To generate this simulated data, it was used coupled stochastic processes that are commonly used in this field (1),
following this stochastic differential equations:$$dx=\mu dt+\sigma d{W}_1$$ $$d\sigma =\alpha (x,\sigma )dt+\beta (x,\sigma )d{W}_2$$ Where $x$ is the return of an invested amount and $\sigma$ is the volatility of an asset, which is also considered as a stochastic variable in this model.

Our goal is to create a well structured open-source dataset with this model, then create a multi-task agent with it's data, to solve problems such as forecasting real stock prices and estimation of the model parameters, to foment the creation of a robust model to use in transfer learning tasks, motivated in the huge success that language models had in this area in recent years.

Referências

1 HESTON, S. L. A closed-form solution for options with stochastic volatility with applications to bond
and currency options. Review of Financial Studies, v. 6, n. 2, p. 327-343, 1993.

Revisão por pares

Paper

Aceito em 20 de jan. de 2021 por Administrador Admin

P_ster_SIFSC_10.pdf