Leakage-Safe Borrower Segmentation Using Deep Clustering and Machine Learning in Large-Scale Lending Data
DOI:
https://doi.org/10.59746/z76aad94Keywords:
Autoencoder, Borrower Segmentation, Deep Clustering, Ex-Ante Credit Risk, Machine Learning, Supervised OperationalizationAbstract
Large-scale lending data often combine pre-origination borrower information with post-origination servicing and outcome fields, which can create data leakage and misleading model performance. This study proposes a leakage-safe borrower-segmentation framework for LendingClub data. The framework constructs an ex-ante analytical matrix, compares direct K-Means/Gaussian Mixture Model (GMM) baselines with deep latent-space clustering using Deep Embedded Clustering (DEC), Improved Deep Embedded Clustering (IDEC), and Deep Clustering Network (DCN), and evaluates supervised operationalization using XGBoost, Logistic Regression (LR), Support Vector Machine (SVM), and Random Forest (RF) . The final matrix contained 2,168,488 observations and 89 features. Conventional clustering showed weak separation, with the best Silhouette (SIL) value of 0.120454, whereas DEC achieved SIL = 0.851837, Davies-Bouldin index (DBI) = 0.179926, and Calinski-Harabasz index (CH) = 1.349271 × 10^6. XGBoost achieved the strongest recoverability performance for the DEC-derived labels. The results support leakage-safe preparation followed by latent deep clustering, while supervised learning is interpreted as operational recoverability of cluster-derived labels rather than direct default prediction.
References
[1] S. H. Goldmann, M. R. Machado, and J. R. Osterrieder, “Advancing credit risk assessment in the retail banking industry: A hybrid approach using time series and supervised learning models,” Data Knowl. Eng., p. 102490, 2025.
[2] T. Miljkovic and P. Wang, “A dimension reduction assisted credit scoring method for big data with categorical features,” Financial Innovation, vol. 11, no. 1, pp. 1–30, 2025, doi: 10.1186/s40854-024-00689-1.
[3] K. Hayat and B. Magnier, “Data Leakage and Deceptive Performance: A Critical Examination of Credit Card Fraud Detection Methodologies,” Mathematics, vol. 13, no. 16, pp. 1–28, 2025, doi: 10.3390/math13162563.
[4] S. R. Lenka, S. K. Bisoy, and R. Priyadarshini, “Multiple optimized ensemble learning for high-dimensional imbalanced credit scoring datasets,” Knowl. Inf. Syst., vol. 66, no. 9, pp. 5429–5457, 2024, doi: 10.1007/s10115-024-02129-z.
[5] X. Yang, H. Xue, Q. Hu, and Y. Zhang, “Design of a full-cycle intelligent risk control system for pre-loan, mid-loan, and post-loan lending: AI-driven closed-loop management of online credit security,” in Proceedings of the 2025 2nd International Conference on Digital Economy and Computer Science, 2025, pp. 1022–1027.
[6] Q. Fan, “Risk Prediction Model of Financial Lending Big Data Leakage Based on Association Rules,” in Smart Innovation, Systems and Technologies, T. Wang, S. Patnaik, W. C. Ho Jack, and M. L. Rocha Varela, Eds., Singapore: Springer Nature Singapore, 2023, pp. 617–629. doi: 10.1007/978-981-19-2768-3_60.
[7] J. P. Noriega, L. A. Rivera, and J. A. Herrera, “Machine learning for credit risk prediction: A systematic literature review,” Data (Basel)., vol. 8, no. 11, p. 169, 2023.
[8] V. Chang, S. Sivakulasingam, H. Wang, S. T. Wong, M. A. Ganatra, and J. Luo, “Credit risk prediction using machine learning and Deep Learning: A study on Credit card customers. Risks, 12 (11), 174,” 2024.
[9] F. Baser, O. Koc, and A. S. Selcuk-Kestel, “Credit risk evaluation using clustering based fuzzy classification method,” Expert Syst. Appl., vol. 223, p. 119882, 2023.
[10] Z. Wang, Q. Li, H. Zhao, and F. Nie, “Simultaneous local clustering and unsupervised feature selection via strong space constraint,” Pattern Recognit., vol. 142, p. 109718, 2023, doi: https://doi.org/10.1016/j.patcog.2023.109718.
[11] W. Lippitt, N. E. Carlson, J. Arbet, T. E. Fingerlin, L. A. Maier, and K. Kechris, “Limitations of clustering with PCA and correlated noise,” J. Stat. Comput. Simul., vol. 94, no. 10, pp. 2291–2319, 2024, doi: 10.1080/00949655.2024.2329976.
[12] T. Li, G. Kou, and Y. Peng, “A new representation learning approach for credit data analysis,” Inf. Sci. (N. Y)., vol. 627, pp. 115–131, 2023, doi: 10.1016/j.ins.2023.01.068.
[13] R. A. Mancisidor, M. Kampffmeyer, K. Aas, and R. Jenssen, “Learning latent representations of bank customers with the variational autoencoder,” Expert Syst. Appl., vol. 164, p. 114020, 2021.
[14] S. Chantamunee, P. Thamrongrat, P. Thanathamathee, K. Chaisriya, and D. N. M. Nizam, “Unsupervised Deep Clustering With Hard Balanced Constraint: Application in Disciplinary-Focused Student Section Formation,” IEEE Access, vol. 12, no. April, pp. 98239–98253, 2024, doi: 10.1109/ACCESS.2024.3423807.
[15] B. V Sánchez Vinces, E. Schubert, A. Zimek, and R. L. F. Cordeiro, “A comparative evaluation of clustering-based outlier detection,” Data Min. Knowl. Discov., vol. 39, no. 2, p. 13, 2025.
[16] B. Lafabregue, J. Weber, P. Gançarski, and G. Forestier, “End-to-end deep representation learning for time series clustering: a comparative study,” Data Min. Knowl. Discov., vol. 36, no. 1, pp. 29–81, 2022, doi: 10.1007/s10618-021-00796-y.
[17] M. Sadeghi and N. Armanfard, “Deep Multirepresentation Learning for Data Clustering,” IEEE Trans. Neural Netw. Learn. Syst., vol. 35, no. 11, pp. 15675–15686, 2024, doi: 10.1109/TNNLS.2023.3289158.
[18] K. Wang, M. Li, J. Cheng, X. Zhou, and G. Li, “Research on personal credit risk evaluation based on XGBoost,” Procedia Comput. Sci., vol. 199, pp. 1128–1135, 2021, doi: 10.1016/j.procs.2022.01.143.
[19] H. Li, “Comparative Study of Personal Credit Default Risk Prediction Based on Different Machine Learning Models,” in ITM Web of Conferences, EDP Sciences, 2025, p. 1016.
[20] X. Huang, “Research on Credit Risk Prediction Technology Based on Deep Learning,” in Proceedings of the International Conference on Image Processing, Machine Learning and Pattern Recognition, 2024, pp. 270–274.
[21] P. Kündig and F. Sigrist, “A Spatio-Temporal Machine Learning Model for Mortgage Credit Risk: Default Probabilities and Loan Portfolios,” Eur. J. Oper. Res., 2025.
[22] M. Gramifar, G. Kabir, and S. A. Khan, “Credit risk assessment using bayesian networks and machine learning approaches,” Advanced Engineering Informatics, vol. 74, p. 104638, 2026.
[23] T. Liu and D. Huang, “Research on credit risk assessment based on machine learning,” in Fifth International Conference on Mechatronics and Computer Technology Engineering (MCTE 2022), SPIE, 2022, pp. 1420–1425.
[24] L. Yu, L. Yu, and K. Yu, “A high-dimensionality-trait-driven learning paradigm for high dimensional credit classification,” Financial Innovation, vol. 7, no. 1, 2021, doi: 10.1186/s40854-021-00249-x.
[25] L. Chen, L. Sun, and C. Zhong, “An integrated interpretation and clustering model based on attribute grouping,” Applied Intelligence, vol. 55, no. 7, p. 599, 2025, doi: 10.1007/s10489-025-06262-2.
[26] J. Bosker, M. Gürtler, and M. Zöllner, “Machine learning-based variable selection for clustered credit risk modeling,” Journal of Business Economics, vol. 95, no. 4, pp. 617–652, 2025, doi: 10.1007/s11573-024-01213-8.
