XGBoost-Based Web Application for Optimizing Credit Evaluation in Peruvian Loan Fintechs

Ysla R.S.G.; Cuba K.A.A.; Alarcon J.S.D. · International Conference on Artificial Intelligence, Computer, Data Sciences, and Applications, ACDSA 2026 (2026) · DOI: 10.1109/acdsa67686.2026.11467742

Resumen

The rapid growth of the fintech sector in Peru has created an urgent need for more accurate and automated credit evaluation systems, as traditional approaches often fail to consider external or alternative data sources. This paper proposes the development of a web application based on the XGBoost algorithm to optimize credit risk prediction in Peruvian fintech companies offering personal loans. The study followed four stages: data collection and preparation, model training, web application construction, and validation. A dataset of 59,682 historical loan applications provided by a Peruvian fintech company was used, after applying data cleaning, normalization, and categorical encoding techniques. The XGBoost model was compared against Random Forest and Logistic Regression using classification metrics such as accuracy, recall, F1-score, and AUC-ROC. Results show that XGBoost achieved an AUC-ROC of 0.965 and an F1score of 0.977, offering a balanced trade-off between accuracy and computational efficiency. When integrated into the web platform, the model reduced the average credit evaluation time by over 60 percent and increased accuracy by 13.2 percentage points compared to the manual process. Overall, the proposed system demonstrates that using XGBoost in a web-based environment significantly improves the automation, consistency, and speed of credit evaluation processes, contributing to more reliable financial decision-making in Peruvian fintechs.

Instituciones: Universidad Peruana de Ciencias Aplicadas