Scholarlink

Financial decision-making heavily relies on price prediction of stock, which leads to the investigation of numerous machine learning strategies for precise future price forecasting. For prediction of stock prices, this research study compares the long short-term memory (LSTM) and gated recurrent unit (GRU) models. The study uses IBM’s historical stock data, preprocesses it, and to assess how well the LSTM and GRU models perform, use of root mean squared error (RMSE) as a statistic to its performance is taken. The models undergo training using consecutive data, and the models’ forecasts are with the real stock prices. To improve model performance, the research also uses methods like data standardization, feature engineering, and sequence generation. The Python Keras package is used to implement the LSTM and GRU models, using its capacity to recognize dependencies of sequential data for long term. The GRU model uses a similar architecture with gated recurrent units; however, the LSTM model has numerous LSTM layers followed by dense layers. The models are assessed on a test dataset after being trained with the Adam optimizer. The study also addresses the usefulness of stock price prediction in financial markets, highlighting its importance in trading methods, risk evaluation, and portfolio management. The results of this experiment show that both GRU and LSTM models compete closely to perform better than each other especially in forecasting stock prices. The RMSE of LSTM model stands at 0.006, whereas the GRU model achieves an RMSE of 0.007. The RMSE values are produced to assess the prediction accuracy. We were able to attain highest accuracy of 99% in our LSTM model. The study also emphasizes how crucial precise stock price prediction is for practical uses such as algorithmic trading, sentiment analysis, and financial risk management. In consideration to all things, this study has advanced knowledge in methods of deep learning for stock price prediction and sheds light upon their usefulness in financial decision-making.

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

LSTM-GRU: Stocks Prediction Using Hybrid Machine Learning Algorithms

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Raghav Chathli,
  • Sachin,
  • Prashant Giridhar Shambharkar

摘要

Financial decision-making heavily relies on price prediction of stock, which leads to the investigation of numerous machine learning strategies for precise future price forecasting. For prediction of stock prices, this research study compares the long short-term memory (LSTM) and gated recurrent unit (GRU) models. The study uses IBM’s historical stock data, preprocesses it, and to assess how well the LSTM and GRU models perform, use of root mean squared error (RMSE) as a statistic to its performance is taken. The models undergo training using consecutive data, and the models’ forecasts are with the real stock prices. To improve model performance, the research also uses methods like data standardization, feature engineering, and sequence generation. The Python Keras package is used to implement the LSTM and GRU models, using its capacity to recognize dependencies of sequential data for long term. The GRU model uses a similar architecture with gated recurrent units; however, the LSTM model has numerous LSTM layers followed by dense layers. The models are assessed on a test dataset after being trained with the Adam optimizer. The study also addresses the usefulness of stock price prediction in financial markets, highlighting its importance in trading methods, risk evaluation, and portfolio management. The results of this experiment show that both GRU and LSTM models compete closely to perform better than each other especially in forecasting stock prices. The RMSE of LSTM model stands at 0.006, whereas the GRU model achieves an RMSE of 0.007. The RMSE values are produced to assess the prediction accuracy. We were able to attain highest accuracy of 99% in our LSTM model. The study also emphasizes how crucial precise stock price prediction is for practical uses such as algorithmic trading, sentiment analysis, and financial risk management. In consideration to all things, this study has advanced knowledge in methods of deep learning for stock price prediction and sheds light upon their usefulness in financial decision-making.