Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predicting Mortgage Default with Machine Learning: AutoML, Class Imbalance, and Leakage Control

Xianghong Hu, Tianning Xu, Ying Chen, Shuai Wang

TL;DR

This work tackles mortgage default prediction under two key challenges: labeling ambiguity and severe class imbalance, compounded by information leakage from temporal data. The authors implement leakage-aware feature selection, a strict dual-dimensional temporal split, and downsampling to create reliable evaluation conditions, then compare logistic regression, tree ensembles, gradient boosting, and AutoML (AutoGluon) on the Fannie Mae loan-performance dataset. AutoGluon achieves the strongest AUROC across downsampling regimes, while leakage control and careful feature processing are shown to be essential to avoid optimistic biases and overfitting. The findings demonstrate that AutoML, when applied with explicit leakage mitigation, offers a practical and effective tool for mortgage risk assessment and decision support.

Abstract

Mortgage default prediction is a core task in financial risk management, and machine learning models are increasingly used to estimate default probabilities and provide interpretable signals for downstream decisions. In real-world mortgage datasets, however, three factors frequently undermine evaluation validity and deployment reliability: ambiguity in default labeling, severe class imbalance, and information leakage arising from temporal structure and post-event variables. We compare multiple machine learning approaches for mortgage default prediction using a real-world loan-level dataset, with emphasis on leakage control and imbalance handling. We employ leakage-aware feature selection, a strict temporal split that constrains both origination and reporting periods, and controlled downsampling of the majority class. Across multiple positive-to-negative ratios, performance remains stable, and an AutoML approach (AutoGluon) achieves the strongest AUROC among the models evaluated. An extended and pedagogical version of this work will appear as a book chapter.

Predicting Mortgage Default with Machine Learning: AutoML, Class Imbalance, and Leakage Control

TL;DR

This work tackles mortgage default prediction under two key challenges: labeling ambiguity and severe class imbalance, compounded by information leakage from temporal data. The authors implement leakage-aware feature selection, a strict dual-dimensional temporal split, and downsampling to create reliable evaluation conditions, then compare logistic regression, tree ensembles, gradient boosting, and AutoML (AutoGluon) on the Fannie Mae loan-performance dataset. AutoGluon achieves the strongest AUROC across downsampling regimes, while leakage control and careful feature processing are shown to be essential to avoid optimistic biases and overfitting. The findings demonstrate that AutoML, when applied with explicit leakage mitigation, offers a practical and effective tool for mortgage risk assessment and decision support.

Abstract

Mortgage default prediction is a core task in financial risk management, and machine learning models are increasingly used to estimate default probabilities and provide interpretable signals for downstream decisions. In real-world mortgage datasets, however, three factors frequently undermine evaluation validity and deployment reliability: ambiguity in default labeling, severe class imbalance, and information leakage arising from temporal structure and post-event variables. We compare multiple machine learning approaches for mortgage default prediction using a real-world loan-level dataset, with emphasis on leakage control and imbalance handling. We employ leakage-aware feature selection, a strict temporal split that constrains both origination and reporting periods, and controlled downsampling of the majority class. Across multiple positive-to-negative ratios, performance remains stable, and an AutoML approach (AutoGluon) achieves the strongest AUROC among the models evaluated. An extended and pedagogical version of this work will appear as a book chapter.
Paper Structure (12 sections, 3 figures, 3 tables)

This paper contains 12 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Dataset and Labeling
  5. Temporal Partitioning and Leakage Mitigation
  6. Feature Selection and Transformation
  7. Downsampling
  8. Models
  9. Experimental Results
  10. Conclusion
  11. Additional Experimental Results
  12. Model Hyperparameters

Figures (3)

  • Figure 1: Temporal partition by ORIG_DATE and ACT_PERIOD.
  • Figure 2: ROC curves at 1:2 positive-to-negative ratio.
  • Figure 3: AutoGluon feature importance at 1:2 ratio.