Research map / Label-efficient and robust learning / Recognition and long-tail
Enhanced Random Forest Algorithms for Partially Monotone Ordinal Classification
Christopher Bartley, Wei Liu, Mark Reynolds
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace XGBoost-style inherited-constraint monotonisation with the proposed MR-RF rule monotonisation | Random Forest trees · Simulated data, in-sample accuracy of globally monotone trees | training accuracy | Improved |
| Use MR-RF-log instead of MR-RF-bay for rule coefficients (ablation) | MR-RF · Simulated data, in-sample accuracy | training accuracy | Improved |
| Use MR-RF-log instead of standard Random Forest (adds monotonicity) | Random Forest · 17 datasets | mean Kappa: +0.025 | Improved |
| Use MR-RF-bay instead of standard Random Forest (adds monotonicity) | Random Forest · 17 datasets | mean Kappa: +0.029 | Improved |
| MR-RF-log versus the globally monotone ICT-RF | MR-RF-log · 17 datasets | MAE: p = 0.169 | No clear effect |
| MR-RF-bay versus the globally monotone ICT-RF | MR-RF-bay · 17 datasets | MAE: p = 0.714 | No clear effect |
| MR-RF-bay versus local approaches (FSD-RF, MID-RF, PM-RF) | MR-RF-bay · 17 datasets | Kappa, F-measure, MAE: p > 0.3 | No clear effect |
| MR-RF-log versus PM-RF | MR-RF-log · 17 datasets | MAE: p = 0.169 | No clear effect |
| Use global monotonicity (MR-RF) instead of local split or penalty constraints | FSD-RF, PM-RF, MID-RF · 17 datasets | MNCmax: 0.9%, 1.6%, 1.8% medians for the local methods | Improved |
| Local constraints on a highly non-monotone dataset | FSD-RF and MID-RF · ERA | MNCmax: 15% | Hurt |
| Use PM-RF instead of FSD-RF or MID-RF on ERA | monotone RF · ERA | MNCmax: 15% -> 5.2% | Improved |
| Use closed-form Bayesian coefficients (MR-RF-bay) instead of constrained logistic regression (ablation) | MR-RF · solve time | computational complexity | Improved |
Closest papers by meaning
- Towards Convergence Rate Analysis of Random Forests for Classification (NeurIPS 2020 · similarity 0.85)
- Comparison-Based Random Forests (ICML 2018 · similarity 0.85)
- Random Forests, Decision Trees, and Categorical Predictors: The "Absent Levels" Problem (JMLR 2018 · similarity 0.85)
- Hierarchical Shrinkage: Improving the accuracy and interpretability of tree-based models. (ICML 2022 · similarity 0.84)
- Generalized Random Forests Using Fixed-Point Trees (ICML 2025 · similarity 0.82)
- DynFrs: An Efficient Framework for Machine Unlearning in Random Forest (ICLR 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
