企業信用評級是資本市場裡投資者對一家公司的重要判斷依據，公開評級結果能夠使企業的違約風險與信用狀況更加公開透明，以降低投資者與經營者之間的資訊不對稱，讓資本市場的交易更公平。

由於傳統的企業信用評級評定過程非常耗費時間與金錢成本，為了即時且快速地因應市場變化作出更新，採用機器學習方法來預測企業信用評級成了近年常被討論的研究課題。

本研究將文本資料(公司管理階層觀點)透過自然語言模型(NLP)處理後匯出文本向量表示，並加入檢索增強生成(RAG)的輸出結果，最後再結合傳統財務數據並匯入極限梯度提升法(XGBoost)進行企業信用評級的預測，來獲得一個兼顧準確率與可解釋力的多元分類模型。

Corporate credit rating serves as a critical reference for investors in the capital market to evaluate a company. Publicly available rating results enhance transparency regarding a company’s default risk and credit status, reducing information asymmetry between investors
and operators, and promoting fairer transactions in the capital market.

Traditional corporate credit rating processes are time-consuming and costly. To respond to market changes promptly and efficiently, applying machine learning methods to predict corporate credit ratings has become a widely discussed research topic in recent years.

This study leverages textual data (MD&A), processed through natural language models (NLP), to generate text vector representations. It incorporates the output of Retrieval-Augmented Generation (RAG) and integrates it with traditional financial data. These inputs are then fed into the Extreme Gradient Boosting (XGBoost) algorithm to predict corporate credit ratings, aiming to develop a multi-class classification model that balances accuracy and interpretability.

[1] Alberti, C., Lee, K., & Collins, M. (2019). A BERT Baseline for the Natural
Questions. arXiv preprint arXiv:1901.08634.
[2] Bengio, Y., Ducharme, R., & Vincent, P. (2003). A Neural Probabilistic Language
Model. Advances in Neural Information Processing Systems 13 (NIPS 2000),50-56.
[3] Bochkay, K., & Levine, Carolyn B. (2019). Using MD&A to Improve Earnings
Forecasts. Journal of Accounting, 34(3), 458–482.
[4] Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data
Mining,785-794.
[5] Choi, J., Suh, Y., & Jung, N. (2020). Predicting Corporate Credit Rating Based on Qualitative Information of MD&A Transformed Using Document Vectorization Techniques.
Data Technologies and Applications, 54(2), 151–168.
[6] Cuconasu, F., Trappolini, G., & Siciliano, F. (2024). The Power of Noise: Redefining Retrieval for RAG Systems. Proceedings of the 47th International ACM SIGIR Conference on
Research and Development in Information Retrieval, 719-729.
[7] Demoulin, N. T. M., & Coussement, D. (2020). Acceptance of Text-Mining Systems: The Signaling Role of Information Quality. Information & Management, 57(1), 78-82.
[8] Durnev, A., & Mangen, C. (2020). The Spillover Effects of MD&A Disclosures for Real Investment: The Role of Industry Competition. Journal of Accounting and Economics, 70(1),
24-32.
[9] Feldman, R., Govindaraj, S., Livnat, J. and Segal, B. (2010). Management’s tone change, post earnings announcement drift and accruals. Review of Accounting Studies, 15(4), 915-953.
[10] Gao, Y., Xiong, Y., & Gao, X. (2024). Retrieval-Augmented Generation for Large
Language Models: A Survey. arXiv preprint arXiv:2312.10997.
[11] Golbayani, P., Florescu, I., & Chatterjee, R. (2020). A Comparative Study of Forecasting Corporate Credit Ratings Using Neural Networks, Support Vector Machines, and Decision
Trees. The North American Journal of Economics and Finance, 54, 77-82.
[12] Golbayani, P., Wang, D., & Florescu, I. (2020). Application of Deep Neural Networks to
Assess Corporate Credit Rating. arXiv preprint arXiv:2003.02334.
[13] Hajek, P. and Michalak, K. (2013).Feature selection in corporate credit rating prediction.
KnowledgeBased Systems, 51(4), 72-84.
[14] Hinton, G., Vinyals, O., & Dean, J. (2015). Distilling the Knowledge in a Neural Network.
arXiv preprint arXiv:1503.02531.
[15] Huang, Z., Chen, H., Hsu, C.J., Chen, W.H. and Wu, S. (2004). Credit rating analysis with support vector machines and neural networks: a market comparative study. Decision Support
Systems, 37(4), 543-558.
[16] Kim, K., & Ahn, H. (2012). A Corporate Credit Rating Model Using Multi-Class Support Vector Machines with an Ordinal Pairwise Partitioning Approach. Computers & Operations
Research, 39(8), 1800-1811.
[17] Lee, Y. (2007). Application of Support Vector Machines to Corporate Credit Rating
Prediction. Expert Systems with Applications, 33(2), 67-74.
[18] Lewis, P., & Perez, E. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. Advances in Neural Information Processing Systems 33 (NeurIPS 2020),105-124.
[19] Li, J., Yuan, Y., & Zhang, Z. (2024). Enhancing LLM Factual Accuracy with RAG to Counter Hallucinations: A Case Study on Domain-Specific Queries in Private Knowledge-
Bases. arXiv preprint arXiv:2403.10446.
[20] Mikolov, T., Chen, K., Corrado, G., & Dean, J. (2013). Efficient Estimation of Word
Representations in Vector Space. arXiv preprint arXiv:1301.3781.
[21] Sanh, V., Debut, L., Chaumond, J., & Wolf, T. (2020). DistilBERT, a Distilled Version of
BERT: Smaller, Faster, Cheaper and Lighter. arXiv preprint arXiv:1910.01108.
[22] Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention Is All You Need. Advances in Neural Information Processing
Systems 30 (NIPS 2017),136-142.
[23] Wang, M., & Ku, H. (2021). Utilizing Historical Data for Corporate Credit Rating
Assessment. Expert Systems With Applications, 165(1),49-57.
[24] Ye, Y., Liu, S., & Li, J. (2008). A Multiclass Machine Learning Approach to Credit Rating
Prediction. IEEE Xplore, 57-61.
[25] Zhang, S., Xu, J., Zhang, Q.J., & Root, D. E. (2016). Parallel matrix neural network training on cluster systems for dynamic FET modeling from large datasets. IEEE Xplore,1-3.