Scholarlink

This paper demonstrates the effectiveness of deep learning algorithms compared to traditional mathematical and machine learning algorithms in forecasting stock price volatility. Specifically, a Long Short-Term Memory (LSTM) network, representing deep learning approaches, is evaluated against ARIMA and GARCH mathematical models, as well as Support Vector Machines and Random Forest machine learning algorithms. The research aims to identify the strengths and limitations of each model in terms of prediction accuracy and efficiency by analyzing daily stock prices and trading volumes of selected companies. The findings suggest that the LSTM model offers significant advantages in capturing the complex and dynamic nature of financial markets, providing more reliable forecasting tools for investors and policymakers. The study also highlights the trade-off between accuracy and computational requirements, concluding that the GARCH model may be a suitable option when accuracy is not the primary factor, and computation speed is of utmost importance.

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

Evaluating Deep Learning Against Mathematical and Machine Learning Models in Stock Price Volatility Forecasting

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Mykola Zlobin,
  • Volodymyr Bazylevych

摘要

This paper demonstrates the effectiveness of deep learning algorithms compared to traditional mathematical and machine learning algorithms in forecasting stock price volatility. Specifically, a Long Short-Term Memory (LSTM) network, representing deep learning approaches, is evaluated against ARIMA and GARCH mathematical models, as well as Support Vector Machines and Random Forest machine learning algorithms. The research aims to identify the strengths and limitations of each model in terms of prediction accuracy and efficiency by analyzing daily stock prices and trading volumes of selected companies. The findings suggest that the LSTM model offers significant advantages in capturing the complex and dynamic nature of financial markets, providing more reliable forecasting tools for investors and policymakers. The study also highlights the trade-off between accuracy and computational requirements, concluding that the GARCH model may be a suitable option when accuracy is not the primary factor, and computation speed is of utmost importance.