CRE-CNN-LSTM-XAI: An Explainable Deep Learning Framework for Risk Assessment of Real Estate Collateralized Financial Products

Ziwen Zhong; Yuheng Huang

doi:10.63313/JCSFT.9065

Authors

Ziwen Zhong Central University of Finance and Economics, Beijing, China Author
Yuheng Huang University of California Berkeley, Berkeley, CA, USA Author

DOI:

https://doi.org/10.63313/JCSFT.9065

Keywords:

Real estate finance, Mortgage-backed securities, Risk assessment, CNN-LSTM, Explainable artificial intelligence, Financial risk management

Abstract

Risk assessment of real estate collateralized financial products, such as mortgage-backed securities (MBS) and real estate investment trusts (REITs), is a core task in banking risk control, structured product issuance, and financial regulation. The strong dependence of these products on real estate price dynamics, interest rate fluctuations, and macroeconomic conditions poses significant challenges to traditional risk evaluation methods that rely on static indicators and linear assumptions. To address these limitations, this paper proposes CRE-CNN-LSTM-XAI, an explainable deep learning framework for AI-driven risk assessment of real estate collateralized financial products. The proposed model integrates a convolutional neural network (CNN) to capture short-term local fluctuations in multivariate real estate and macroeconomic time series with a long short-term memory (LSTM) network to model long-term dependencies and cyclical market trends. Based on the learned latent risk representations, the framework predicts the default probability of collateralized real estate financial assets. To enhance transparency and practical applicability, a SHAP-based explainable artificial intelligence module is incorporated to quantify the contribution of key risk factors, supporting regulatory compliance and decision-making in financial institutions. Experiments conducted on real estate market and mortgage-related datasets demonstrate that CRE-CNN-LSTM-XAI consistently outperforms benchmark models. The proposed framework achieves an AUC of 0.87, compared with 0.82 for a standalone LSTM and 0.79 for XGBoost. These results indicate that CRE-CNN-LSTM-XAI effectively captures complex temporal risk patterns while providing meaningful economic interpretability, making it well suited for practical applications in bank risk management and mortgage-backed securities issuance.

References

[1] Deng Y, Quigley J M, Van Order R. Mortgage terminations, heterogeneity and the exercise of mortgage options[J]. Econometrica, 2000, 68(2): 275-307.

[2] Kau J B, Keenan D C, Muller III W J, et al. The valuation at origination of fixed-rate mortgages with default and prepayment[J]. The Journal of Real Estate Finance and Economics, 1995, 11(1): 5-36.

[3] Gerardi K, Shapiro A H, Willen P S. Subprime outcomes: Risky mortgages, homeownership experiences, and foreclosures[R]. Working papers, 2008.

[4] Breiman L. Random forests[J]. Machine learning, 2001, 45(1): 5-32.

[5] Chen T, Guestrin C. Xgboost: A scalable tree boosting system[C]//Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining. 2016: 785-794.

[6] Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural computation, 1997, 9(8): 1735-1780.

[7] Bao W, Yue J, Rao Y. A deep learning framework for financial time series using stacked autoencoders and long-short term memory[J]. PloS one, 2017, 12(7): e0180944.

[8] Sirignano J, Cont R. Universal features of price formation in financial markets: perspectives from deep learning[M]//Machine learning and AI in finance. Routledge, 2021: 5-15.

[9] Ge C. A LSTM and graph CNN combined network for community house price forecasting[C]//2019 20th IEEE International Conference on Mobile Data Management (MDM). IEEE, 2019: 393-394.

[10] Xu S, Jiang L, Gu B. Design and Validation of a Smart Neuromorphic System Architecture for Algorithmic Trading[C]//Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025: 127-136.

[11] Fan P, Li H, Hu M. Profit-Oriented Production and Pricing Optimization for Manufacturing Enterprises Using Proximal Policy Optimization[J]. Economics and Management Innovation, 2026, 3(2): 8-17.

[12] Lundberg S M, Lee S I. A unified approach to interpreting model predictions[J]. Advances in neural information processing systems, 2017, 30.

[13] Patidar D, Fallah M H, Hemalatha K, et al. Explainable AI Using SHAP for Transparent and Interpretable Decision-Making in Financial Risk Assessment and Credit Scoring[C]//2025 3rd International Conference on Cyber Resilience (ICCR). IEEE, 2025: 1-7.

[14] Molnar C. Interpretable machine learning[M]. Lulu. com, 2020.

[15] Li, Zhenghang, Xinyu Li, and Xinning Lin. "Design and Implementation of a Platform for Business Intelligence Knowledge Mining and Graph Construction Based on Deep Learning." Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025.

[16] Luo R H, Wang N, Zhu X. Fraud detection and risk assessment of online payment transactions on e-commerce platforms based on llm and gcn frameworks[J]. arXiv preprint arXiv:2509.09928, 2025.

[17] Li T, Li H, Zhou Y. E-commerce Sentiment Analysis Using Fine-tuned LLaMA3 Models: A QLoRA-based Approach[J]. Journal of Technology Innovation and Engineering, 2025, 1(4).

[18] Luo R, Hu J, Sun Q. Group Anomaly Detection and Risk Control of Commodity Sales Volume Data Based on LSTM-VAE Framework[J]. Journal of Computer, Signal, and System Research, 2025, 2(7): 48-57.