Scholarlink

In this paper, we utilize fuzzy logic in combination with a Many-to-Many Long Short-Term Memory (MTM LSTM) and a Multilayer Perceptron (MLP) neural network to forecast stock prices. This hybrid approach offers a useful tool for investors, helping them make better-informed decisions and enhance their portfolio management strategies. We tested the model using monthly stock closing prices from the Vietnam Stock Exchange, focusing on predictions for up to six months into the future. The experimental findings highlight the model’s high accuracy, as indicated by low error metrics such as Mean Squared Error (MSE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE), with predicted values closely aligning with actual market data. Furthermore, the strong R-squared values demonstrate the model’s capability to capture data variability effectively, reinforcing its reliability for stock price forecasting. For a one-month forecast, we obtained MSE, RMSE, MAE, and R-squared values of 2.89, 1.70, 1.20, and 99.22%, respectively.

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

Advanced Integration of Fuzzy MTM LSTM and MLP Models for Forecasting Stock Prices

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Phuc Tan Huynh,
  • Trang Hong Son,
  • Khoa Dang Vo,
  • Nguyen Huynh-Tuong,
  • Ameur Soukhal

摘要

In this paper, we utilize fuzzy logic in combination with a Many-to-Many Long Short-Term Memory (MTM LSTM) and a Multilayer Perceptron (MLP) neural network to forecast stock prices. This hybrid approach offers a useful tool for investors, helping them make better-informed decisions and enhance their portfolio management strategies. We tested the model using monthly stock closing prices from the Vietnam Stock Exchange, focusing on predictions for up to six months into the future. The experimental findings highlight the model’s high accuracy, as indicated by low error metrics such as Mean Squared Error (MSE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE), with predicted values closely aligning with actual market data. Furthermore, the strong R-squared values demonstrate the model’s capability to capture data variability effectively, reinforcing its reliability for stock price forecasting. For a one-month forecast, we obtained MSE, RMSE, MAE, and R-squared values of 2.89, 1.70, 1.20, and 99.22%, respectively.