Jingzhi Yin
The Department of Mathematics, Columbia University in the City of New York, New York, NY 10027, USA.
*Corresponding author: Jingzhi Yin
Abstract
Against the backdrop of a complex macroeconomic environment and intensified fluctuations in credit risk cycles, spreads in the CLO secondary market exhibit increasingly pronounced nonlinear and sentiment-driven characteristics. Traditional forecasting approaches based on fundamentals and market variables face limitations in capturing short-term spread dynamics. This study introduces sentiment factors derived from the RoBERTa pre-trained language model, extracting investor sentiment signals from unstructured textual information and integrating them with CLO secondary market spread volatility to construct a sentiment-driven forecasting model. Through a systematic process of sentiment factor quantification and feature engineering, textual sentiment information is transformed into numerical variables suitable for prediction, followed by model training and optimization. Empirical results show that sentiment factors significantly enhance both the explanatory power and predictive accuracy of spread volatility, with the model demonstrating strong robustness across different testing conditions. These findings provide empirical evidence on the role of sentiment information in the pricing of structured credit products and offer new analytical perspectives for risk management and investment decision-making in the CLO market.
References
[1] Kumar P, Ganapathy SM. An Analysis of Marketing Channels and Price Spread of Chrysanthemum in Chikkaballapura District of Karnataka. Asian J Agric Ext Econ Soc. 2023;41(5):93-98.
[2] Luo M, Liu S, Zhu L, et al. Analysis of a super-transmission of SARS-CoV-2 omicron variant BA.5.2 in the outdoor night market. Front Public Health. 2023;11:1153303.
[3] Jiang C, Zhang Y, Han Y. Do Bond Investors Care about Margin Trading and Securities lending?:—An Empirical Study Based on Secondary Market of Medium Term Notes. Procedia Computer Science. 2022 Jan 1;199:613-20.
[4] Wang S, Liu Q, Hu Y, et al. Public Opinion Evolution Based on the Two-Dimensional Theory of Emotion and Top2Vec-RoBERTa. Symmetry. 2025;17(2):190.
[5] Ramalingaswamy C, Khaja H, Ilaiah K, et al. Sentiment classification with modified RoBERTa and recurrent neural networks. Multimed Tools Appl. 2023;83(10):29399-417.
[6] Redouane E, Yoshio N. Fire-sale risk in the leveraged loan market. J Financ Econ. 2022;146(3):1120-47.
[7] Maranon Issues Eighth Middle Market CLO. Manuf Close-Up. 2021.
[8] Dennis V, Mike N, Vivian BV. Security design and credit rating risk in the CLO market. J Int Financ Mark Inst Money. 2021;72:101290.
[9] Lak JA, Boostani R, Alenizi AF, et al. RoBERTa, ResNeXt and BiLSTM with self-attention: The ultimate trio for customer sentiment analysis. Appl Soft Comput. 2024;164:112018.
How to cite this paper
Research on a CLO Secondary Market Spread Volatility Prediction Model Based on RoBERTa Sentiment Factors
How to cite this paper: Jingzhi Yin. (2026) Research on a CLO Secondary Market Spread Volatility Prediction Model Based on RoBERTa Sentiment Factors. Advances in Computer and Communication, 7(1), 38-42.
DOI: http://dx.doi.org/10.26855/acc.2026.03.005