Reproduce “Gradient Boosting Survival Tree with Applications in Credit Scoring” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Miaojun Bai, Yan Zheng, Yun Shen
Researchers from 360 Finance Inc. developed Gradient Boosting Survival Tree (GBST), a model that predicts both the probability and timing of loan default, specifically designed for the high-dimensional and heterogeneous data of the Chinese consumer finance market. GBST consistently outperformed traditional survival models and classification methods on large-scale proprietary and public datasets, demonstrating superior discriminative power in predicting time-to-default.
Key insights
- Combining gradient boosting with likelihood-optimized survival trees provides a robust framework for time-to-event prediction, addressing limitations of prior models.
- Modeling the probability distribution of time-to-default, rather than just binary classification, enables more granular and dynamic credit risk management.
- The model is specifically designed to handle the high-dimensional, noisy, sparse, and heterogeneous characteristics of real-world industrial financial data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.