Scholarlink

Financial time series forecasting is a crucial task for financial institutions, investors, and academics. Indeed, financial markets grow increasingly interdependent and highly responsive to both economic and non-economic events. The ability of market participants to accurately manage market spillover determines how well they perform. The goal of implementing machine learning algorithms is to improve the forecasting process’s quality. The architecture of the RNNS family is designed to be optimal for time series forecasting. In this study, we compare three popular machine learning algorithms: CNN, LSTM, and GRU to predict the FTSE CSE index returns. The results of our model comparison indicate that the GRU model outperforms the others, providing more realistic and reliable forecasts based on MAE and RMSE metrics. While all models visually follow the original data trends closely, GRU demonstrates a superior performance in terms of accuracy, as evidenced by the error metrics.

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 FTSE CSE: A Neural Network Model Comparison of GRU, 1D-CNN and LSTM Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ayman Mazzouz,
  • Abdelali Elmoufidi

摘要

Financial time series forecasting is a crucial task for financial institutions, investors, and academics. Indeed, financial markets grow increasingly interdependent and highly responsive to both economic and non-economic events. The ability of market participants to accurately manage market spillover determines how well they perform. The goal of implementing machine learning algorithms is to improve the forecasting process’s quality. The architecture of the RNNS family is designed to be optimal for time series forecasting. In this study, we compare three popular machine learning algorithms: CNN, LSTM, and GRU to predict the FTSE CSE index returns. The results of our model comparison indicate that the GRU model outperforms the others, providing more realistic and reliable forecasts based on MAE and RMSE metrics. While all models visually follow the original data trends closely, GRU demonstrates a superior performance in terms of accuracy, as evidenced by the error metrics.