A Hybrid CNN–BiGRU Model with Grey Wolf Optimization and LightGBM for Stock Price Prediction

• Fatima A. MUSA

  Author

• Abdulmajid B. UMAR

  Author

• Abba M. BALA

  Author

Accurate stock price prediction remains a challenging task due to the non-linearity, volatility, and noisy nature of financial time-series data. Traditional statistical models such as ARIMA are limited in capturing complex market dynamics, “while standalone deep learning models often suffer from overfitting and suboptimal hyperparameter selection”. This paper proposes a hybrid stock price prediction framework that integrates Convolutional Neural Networks (CNN) for feature extraction and Bidirectional Gated Recurrent Units (BiGRU) for temporal dependency modeling. The Grey Wolf Optimizer (GWO) is employed for automated hyperparameter optimization, while Light Gradient Boosting Machine (LightGBM) is used as the final regression layer to enhance generalization. ARIMA is applied as a preprocessing technique to remove linear components and ensure stationarity. Experimental results show that the proposed model significantly outperforms the baseline ARIMA + Attention-CNN-BiLSTM + XGBoost model, achieving a high explanatory power with an R² value of 0.93903, alongside notable reductions in MAE, MSE, and RMSE. These results demonstrate the effectiveness of combining deep learning architectures, metaheuristic optimization, and ensemble learning for accurate and reliable stock price prediction.

• FJET_21_95_95

Copyright (c) 2026 FUDMA Journal of Engineering and Technology

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.