Research on Financial Credit Risk Assessment Model Based on WOE Encoding and Ensemble Learning

Aiming at the limitations of traditional logistic regression in handling high dimensional nonlinear data and class imbalance in financial credit risk assessment, this paper uses the GiveMeSomeCredit dataset. Missing monthly income values were filled using random forest, skewed features were corrected by log quantile transformation, expanding features from 4 dimensions to 27 nonlinear features were transformed using WOE encoding, and an ensemble model was constructed combining logistic regression, random forest, and gradient boosting machine, while class imbalance of 1 14 was handled by under sampling and cost sensitive learning. Experiments show that the GBM model is optimal, improving 9.4 over baseline logistic regression and saving 390,000 yuan annually in bad debt costs WOE encoded logistic regression maintains full interpretability and meets regulatory requirements. The study provides support for risk control decision making in financial institutions. Chang Xinyuan "Research on Financial Credit Risk Assessment Model Based on WOE Encoding and Ensemble Learning" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-9 | Issue-5 , October 2025, URL: https://www.ijtsrd.com/papers/ijtsrd97600.pdf Paper URL: https://www.ijtsrd.com/economics/financial-economics/97600/research-on-financial-credit-risk-assessment-model-based-on-woe-encoding-and-ensemble-learning/chang-xinyuan