A Novel Ensemble Learning Framework Based on News Sentiment Enhancement and Multi-objective Optimizer for Carbon Price Forecasting

Yujie Chen, Mingyao Jin, Zheyu Zhou and Zhirui Tian ()
Additional contact information
Yujie Chen: Dongbei University of Finance and Economics
Mingyao Jin: Dongbei University of Finance and Economics
Zheyu Zhou: Dongbei University of Finance and Economics
Zhirui Tian: The Chinese University of Hong Kong

Computational Economics, 2025, vol. 66, issue 5, No 5, 3709-3733

Abstract: Abstract Carbon price forecasting is crucial for decision-makers, yet it remains a challenging task due to the complex interplay of supply–demand dynamics and the influence of news texts. Existing models predominantly rely on historical data, overlooking the impact of news texts. While some studies enhance prediction accuracy by linearly combining the forecasting results of multiple models using multi-objective optimization algorithms, they neglect the selection process on the Pareto frontier. To address these issues, this paper introduces an ensemble learning framework based on news sentiment enhancement and multi-objective optimizer. In the data preprocessing module based on data denoising and news sentiment enhancement, we utilize successive variational mode decomposition (SVMD) for data denoising, hampel identifier (HI) for outlier removal, and we use latent dirichlet allocation (LDA) to obtain the document-topic matrix of relevant news texts at each time point as input features. in the ensemble learning module, we transition from the football team training algorithm (FTTA) to the multi-objective optimization algorithm (MOFTTA), which allows us to optimize and assign weights to individual forecasting results from the model pool, integrating these weighted forecasts to produce the final forecasting results. In the Pareto Frontier Shrinkage module, using a knee point strategy, we select optimal solutions at the Pareto frontier to balance trade-offs among different objective functions, utilizing knee points derived from knee point identification based on trade-off utility (KPITU) as the optimal solution set. Experiments show that this framework significantly enhances the accuracy and stability of forecasts, outperforming single AI methods.

Keywords: Carbon price forecasting; Ensemble Learning; Multi-objective optimizer; LDA; Pareto front shrinking strategy (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s10614-024-10828-6 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:kap:compec:v:66:y:2025:i:5:d:10.1007_s10614-024-10828-6

Ordering information: This journal article can be ordered from
http://www.springer. ... ry/journal/10614/PS2

DOI: 10.1007/s10614-024-10828-6

Access Statistics for this article

Computational Economics is currently edited by Hans Amman

More articles in Computational Economics from Springer, Society for Computational Economics Contact information at EDIRC.
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().