Forex prophetcan machine learning turn predictions into profits?

Resumo

The forecasting of exchange rates is an important and challenging task for economic agents, given the nonlinearity and instability of models for this asset. In this dissertation, we compare different statistical models and test their forecasting ability for exchange rates. More specifically, we use two datasets: one with financial data at a weekly frequency and another with macroeconomic data at a monthly frequency, covering the period from January 2006 to March 2023. We are particularly interested in determining whether we can predict the correct direction of exchange rate movements to develop a long/short strategy that can be replicated in real life. The results indicate that only a select class of models can achieve consistent profits throughout our sample, but their performance varies considerably depending on the dataset and the period analyzed.

Ficha do documento

Tipo

Dissertação

Ano

2025

Instituição

Fundação Getulio Vargas

Fonte

Repositório da FGV

Idioma

Inglês

Acesso

Acesso aberto

Identificador

oai:repositorio.fgv.br:10438/36894

Temas

Economia

Palavras-chave

CâmbioPrevisão econômicaAprendizado do computadorAnálise de séries temporaisExchange rateForecastingMachine learningTime seriesEconomia