Scholarlink

<p>Accurate time series forecasting is crucial in financial markets, where prediction errors can lead to significant losses. Traditional models often struggle with complex, non-linear patterns, prompting the exploration of hybrid approaches that combine statistical methods with machine learning. This study introduces the ES-LSTM model, a hybrid approach combining Exponential Smoothing (ES) and Long Short-Term Memory (LSTM) networks for improved time series forecasting, particularly in stock market predictions. ES reduces noise for short-term forecasts, while LSTM captures long-term patterns. The results demonstrate that ES-LSTM outperforms traditional models like SARIMAX and Prophet in prediction accuracy. The ES-LSTM model achieves high <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\hbox {R}^{2}\)</EquationSource> </InlineEquation> scores, for example 0.9712 for the S&amp;P 500, and reduces prediction errors. The study uses diverse datasets, spanning different regions, sectors, and market caps to improve generalization and reduce overfitting. The findings suggest that the hybrid approach leverages the strengths of both ES and LSTM, offering improved predictive accuracy and robustness when applied to real-world financial data.</p>

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

ES-LSTM: a hybrid model for accurate time series forecasting in financial markets

  • 期刊论文

  • 发表日期：

    2026年1月19日

查看全文
  • Vaibhav Gagneja,
  • Mayank Gupta,
  • Sanjay Batish,
  • Poonam Saini,
  • Sudesh Rani

摘要

Accurate time series forecasting is crucial in financial markets, where prediction errors can lead to significant losses. Traditional models often struggle with complex, non-linear patterns, prompting the exploration of hybrid approaches that combine statistical methods with machine learning. This study introduces the ES-LSTM model, a hybrid approach combining Exponential Smoothing (ES) and Long Short-Term Memory (LSTM) networks for improved time series forecasting, particularly in stock market predictions. ES reduces noise for short-term forecasts, while LSTM captures long-term patterns. The results demonstrate that ES-LSTM outperforms traditional models like SARIMAX and Prophet in prediction accuracy. The ES-LSTM model achieves high \(\hbox {R}^{2}\) scores, for example 0.9712 for the S&P 500, and reduces prediction errors. The study uses diverse datasets, spanning different regions, sectors, and market caps to improve generalization and reduce overfitting. The findings suggest that the hybrid approach leverages the strengths of both ES and LSTM, offering improved predictive accuracy and robustness when applied to real-world financial data.