Scholarlink

Stock price prediction is vital in financial decision-making by reducing investment risks. This study introduces an ensemble model enhanced with Bagging involving hidden Markov to improve model prediction accuracy and robustness. The proposed method captures temporal dependencies in multivariate stock data using a similarity-based approach in the log-likelihood space across sliding windows. By aggregating multiple HMM learners, the ensemble reduces overfitting and enhances generalization. We assess the model using historical data from Apple (AAPL), Microsoft (MSFT), and Nvidia (NVDA), and benchmark it against several baseline models, including Random Forest (RF), Support Vector Regression (SVM), LSTM, GRU, LightGBM, CatBoost, and XGBoost. Performance is evaluated using metrics commonly used in price prediction. The proposed model consistently achieves superior predictive performance across all stocks and metrics, highlighting its effectiveness in sequential modeling patterns in financial time series.

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

Short -Term Stock Price Prediction Using an Ensemble Hidden Markov Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Thi Viet Huong Nguyen,
  • Dinh Thuan Nguyen

摘要

Stock price prediction is vital in financial decision-making by reducing investment risks. This study introduces an ensemble model enhanced with Bagging involving hidden Markov to improve model prediction accuracy and robustness. The proposed method captures temporal dependencies in multivariate stock data using a similarity-based approach in the log-likelihood space across sliding windows. By aggregating multiple HMM learners, the ensemble reduces overfitting and enhances generalization. We assess the model using historical data from Apple (AAPL), Microsoft (MSFT), and Nvidia (NVDA), and benchmark it against several baseline models, including Random Forest (RF), Support Vector Regression (SVM), LSTM, GRU, LightGBM, CatBoost, and XGBoost. Performance is evaluated using metrics commonly used in price prediction. The proposed model consistently achieves superior predictive performance across all stocks and metrics, highlighting its effectiveness in sequential modeling patterns in financial time series.