Prediksi Keberhasilan Restrukturisasi Kredit Dengan Pendekatan Machine Learning (Model XGboost): Studi Kasus Pada Debitur Korporasi Dan Komersial di Bank BCA
Rony Primasatria, Bowo Setiyono, S.E. , M.Com. , Ph.D. , CFP. , CACP.
2025 | Tesis | S2 MANAJEMEN (MM) JAKARTA
Credit risk management, particularly regarding restructuring failure (re-default), requires more objective predictive tools than manual and subjective methods. This study develops and validates a machine learning model based on the Extreme Gradient Boosting (XGBoost) algorithm to predict restructuring success for Corporate and Commercial debtor segments, using internal data from PT Bank Central Asia Tbk. from 2019 to 2024.
The results show that the developed XGBoost model is highly accurate and reliable, with Area Under the Curve (AUC) values consistently above 0.96 across all test scenarios. The most significant finding is the fundamental duality in the “risk DNA” between the two segments: (1) Corporate Segment: Risk is deterministic and depends on historical records. Restructuring success is almost entirely determined by one dominant factor: a history of Non-Performing Loans. The model for this segment demonstrates near-perfect performance (AUC = 1.0) and remains highly stable, regardless of the training data horizon. (2) Commercial Segment: Risk is probabilistic, dynamic, and multifactorial. Success is determined by a combination of recent factors such as payment behavior, financial health (e.g., Debt to Equity Ratio, sales), and transactional activity. This model is highly sensitive to data recency, achieving the best performance when trained on the most recent 12 months of data.
This study concludes that the XGBoost model effectively quantifies this risk duality. The managerial implication is a recommendation to implement a dual-track decision support system: (1) For Corporate debtors, the model serves as an early warning system, indicating that definitive resolution may be more effective than repeated restructuring; (2) For Commercial debtors, the model acts as a dynamic diagnostic tool requiring continuous monitoring and periodic retraining
Kata Kunci : Restrukturisasi Kredit, Manajemen Risiko Kredit, Machine Learning, XGBoost, Analisis Prediktif, Sistem Peringatan Dini