Paper Contents
Abstract
This paper presents a forecasting framework that integrates Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and ensemble learning models for stock market prediction. Six multi-sector stocksSalesforce, Alibaba, Disney, Electronic Arts, Netflix, and AMDwere analysed to evaluate robustness across different market dynamics. CEEMDAN was applied to decompose raw closing prices into intrinsic mode functions, enhancing signal clarity and reducing noise. The most informative components were selected and used to train Random Forest, LightGBM, and XGBoost regressors. Performance was assessed using RMSE, MAE, and directional accuracy. Results show that Random Forest achieved consistently strong outcomes across most datasets, with Electronic Arts reaching 99.79% accuracy, while LightGBM and XGBoost performed best on Netflix (97.09%) and AMD (79.45%), respectively. These findings confirm that CEEMDAN-driven preprocessing, coupled with ensemble models, provides a reliable, interpretable, and computationally efficient alternative to deep learning approaches for multi-sector financial forecasting.
Copyright
Copyright © 2025 SINDHU S. This is an open access article distributed under the Creative Commons Attribution License.
