Scholarlink

This paper evaluates the performance of Long Short-Term Memory (LSTM) models in forecasting the USD/COP exchange rate using daily financial and macroeconomic data from 2016 to 2020. The model includes 24 explanatory variables grouped into six financial categories. Results show that LSTM achieves a 6.9% forecast error, outperforming traditional ARIMA models (10.9%). These findings highlight the advantages of deep learning for capturing nonlinear dependencies and managing high-dimensional data in volatile emerging markets.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting the USD/COP Exchange Rate Using LSTM Networks: A Comparison with ARIMA Models

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Raul Romero,
  • Diego Leon,
  • Javier Sandoval,
  • German Hernandez,
  • Carlos Zapata

摘要

This paper evaluates the performance of Long Short-Term Memory (LSTM) models in forecasting the USD/COP exchange rate using daily financial and macroeconomic data from 2016 to 2020. The model includes 24 explanatory variables grouped into six financial categories. Results show that LSTM achieves a 6.9% forecast error, outperforming traditional ARIMA models (10.9%). These findings highlight the advantages of deep learning for capturing nonlinear dependencies and managing high-dimensional data in volatile emerging markets.