News Text Embeddings and Clean Energy ETF Return Direction Prediction: An Empirical Study Based on BERT-PCA and LightGBM

Weiliang Liu

doi:10.47297/wspdecWSP2515-797314.20261001

PDF
Export
Share
Collection
Album

News Text Embeddings and Clean Energy ETF Return Direction Prediction: An Empirical Study Based on BERT-PCA and LightGBM

Volume 10, Issue 1

Views: 0 Downloads: 0

News Text Embeddings and Clean Energy ETF Return Direction Prediction: An Empirical Study Based on BERT-PCA and LightGBM
News Text Embeddings and Clean Energy ETF Return Direction Prediction: An Empirical Study Based on BERT-PCA and LightGBM
Universiti Putra Malaysia, Selangor, 47300, Malaysia
Vol. 10, Issue 1, Pages: 67-71(2026)
Published：30 January 2026
DOI：10.47297/wspdecWSP2515-797314.20261001
Accepted：

Scan QR Code

Liu Weiliang.News Text Embeddings and Clean Energy ETF Return Direction Prediction: An Empirical Study Based on BERT-PCA and LightGBM[J].Development Economics of China,2026,10(01):67-71.
DOI：

Liu Weiliang.News Text Embeddings and Clean Energy ETF Return Direction Prediction: An Empirical Study Based on BERT-PCA and LightGBM[J].Development Economics of China,2026,10(01):67-71. DOI： 10.47297/wspdecWSP2515-797314.20261001.

摘要

Abstract

This study investigates whether deep semantic information from financial news headlines provides incremental predictive value for the weekly price direction of the iShares Global Clean Energy ETF (ICLN). Using 189 weeks of data from June 2020 to March 2024

we construct a predictive framework combining BERT embeddings

PCA dimensionality reduction

and a LightGBM classifier

evaluated under a rigorous rolling window validation protocol. Empirical results reveal that although textual factors dominate feature importance—contributing 92.2% of total gain—the full model achieves an out-of-sample AUC of only 0.5359 and a recall of just 11% for upward weeks

failing to generate effective trading signals. These findings indicate that

under the current sample and model specifications

news headline semantics do not provide stable predictive power for ICLN's weekly direction. Methodologically

this paper contributes a transparent pipeline for transforming unstructured news into structured predictors and establishes a rigorous evaluation framework suitable for small-sample financial time series

offering a valuable reference for empirical research in financial natural language processing.

关键词

Keywords

references

Jiang F. , Meng L. , & Tang G . ( 2021 ). Media Textual Sentiment and Stock Return Prediction . Economics (Quarterly) , ( 04 ), 1323 - 1344 .

Jiang F. , Liu Y. , & Meng L . ( 2024 ). Large Language Models, Textual Sentiment, and Financial Markets . Management World , 8 .

Jiang F. , Tang G. , & Ma T . ( 2024 ). Machine Learning for Finance and Financial Management . Beijing : China Machine Press .

Jiang F. , Tang G. , Chen H. , & Zhu L . ( 2025 ). Asset Pricing with Investor Expectation Coordination . Management World , 12 .

Araci D . ( 2019 ). FinBERT: Financial sentiment analysis with pre-trained language models . arXiv: 1908.10063 .

Baker S.R. , Bloom N. , & Davis S . J . ( 2016 ). Measuring economic policy uncertainty. The Quarterly Journal of Economics , 131 ( 4 ), 1593 - 1636 .

Devlin J. , et al . ( 2019 ). BERT: Pre-training of deep bidirectional transformers for language understanding . In *NAACL-HLT 2019 *.

Engelberg J.E. , & Parsons C . A . ( 2011 ). The causal impact of media in financial markets. The Journal of Finance , 66 ( 1 ), 67 - 97 .

Gu S. , Kelly B. , & Xiu D . ( 2020 ). Empirical asset pricing via machine learning . The Review of Financial Studies , 33 ( 5 ), 2223 - 2273 .

Ke G. , et al . ( 2017 ). LightGBM: A highly efficient gradient boosting decision tree. In NeurIPS 2017 .

Loughran T. , & McDonald B . ( 2011 ). When is a liability not a liability? The Journal of Finance , 66 ( 1 ), 35 - 65 .

Loughran T. , & McDonald B . ( 2020 ). Textual analysis in accounting and finance . The Journal of Finance , 75 ( 3 ), 1187 - 1230 .

Tetlock P . C . ( 2007 ). Giving content to investor sentiment. The Journal of Finance , 62 ( 3 ), 1139 - 1168 .

Alert me when the article has been cited

Submit

No data

Related Author

No data

Related Institution

No data

⁰