A Hybrid Clustering-Based Approach for Renewal Opportunity Prioritization in Payroll-Deducted Loan Portfolios
Resumo
The identification of renewal opportunities in payroll-deducted loan portfolios is an important task for financial institutions, as it helps sales teams prioritize contracts with greater operational potential for refinancing. In practice, however, historical labels indicating effective renewal conversions are often unavailable, inconsistent, or unreliable, limiting the direct use of traditional supervised learning approaches. This paper proposes a hybrid data mining approach for renewal opportunity prioritization in payroll-deducted loan portfolios. The method integrates information from loan contracts, payroll margin records, and liquidation history to build an analytical dataset. K-Means clustering is applied to identify contractual profiles, and the cluster with the highest operational adherence to renewal opportunities is used to generate pseudo-labels. Based on these pseudo-labels, a Random Forest model is used as an auxiliary supervised model to produce an operational prioritization score. Experiments were conducted as a pilot study using a real-world dataset with 739 contractual instances from a small municipal agreement. The results indicate that the selected target cluster is characterized by intermediate remaining term, relevant paid percentage, and low available margin. The proposed approach does not estimate actual renewal conversion probabilities, but provides an interpretable decision-support artifact for prioritizing renewal opportunities in scenarios where explicit conversion labels are not available.
Referências
Ayari, H., Guetari, R., and Kraïem, N. Machine learning powered financial credit scoring: a systematic literature review. Artificial Intelligence Review 59 (13), 2026. Published online: 18 November 2025.
Banco Central do Brasil. Empréstimo consignado: características, acesso e uso. Relatório de Cidadania Financeira, 2018. Acesso em: 28 jun. 2026.
Bao, W., Ning, L., and Yue, K. Integration of unsupervised and supervised machine learning algorithms for credit risk assessment. Expert Systems with Applications vol. 128, pp. 301–315, 2019.
Bosker, J., Gürtler, M., and Zöllner, M. Machine learning-based variable selection for clustered credit risk modeling. Journal of Business Economics vol. 95, pp. 617–652, 2025. Published online: 03 December 2024.
Boughaci, D., Alkhawaldeh, A. A. K., Jaber, J. J., and Hamadneh, N. Classification with segmentation for credit scoring and bankruptcy prediction. Empirical Economics 61 (3): 1281–1309, 2021.
Coelho, C. A., De Mello, J. M. P., and Funchal, B. The brazilian payroll lending experiment. The Review of Economics and Statistics 94 (4): 925–934, 2012.
Fredriksson, T., Bosch, J., and Olsson, H. H. An empirical evaluation of deep semi-supervised learning. International Journal of Data Science and Analytics 20 (4): 4127–4148, 2025.
Gomes, R. d. S., Oliveira, E. R. d., Santos, G. C. d., and Gonçalves, R. R. O ambiente socioeconômico influencia o uso de crédito consignado? REDECA, Revista Eletrônica do Departamento de Ciências Contábeis & Departamento de Atuária e Métodos Quantitativos vol. 11, pp. e66978, 2024.
He, C. and Ding, C. H. Q. A novel classification algorithm for customer churn prediction based on hybrid ensemble-fusion model. Scientific Reports 14 (20179), 2024.
Pinheiro, M. A. P. Empréstimos bancários consignados de duas ou mais instituições financeiras: uma perspectiva à luz do direito do consumidor como direito fundamental. Revista da Defensoria Pública da União (13): 21–33, 2020.
Schuh, A. B., Coronel, D. A., and Bender Filho, R. Payroll loans and its relationship with the aggregate economic activity (2004–2014). RAM. Revista de Administração Mackenzie 18 (1): 148–173, 2017.
Tabianan, K., Velu, S., and Ravi, V. K-means clustering approach for intelligent customer segmentation using customer purchase behavior data. Sustainability 14 (12): 7243, 2022.
van Engelen, J. E. and Hoos, H. H. A survey on semi-supervised learning. Machine Learning vol. 109, pp. 373–440, 2020.
