Research map / Probabilistic and scientific modeling / Molecules and proteins
[Re] FOCUS: Flexible Optimizable Counterfactual Explanations for Tree Ensembles
Kyosuke Morita
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 |
|---|---|---|---|
| Use FOCUS instead of Feature Tweaking, Euclidean distance on Wine Quality AdaBoost | tree ensemble counterfactual search · Wine Quality AB | d_mean (Euclidean): 0.267 -> 0.188 | Improved |
| Use FOCUS instead of Feature Tweaking, Euclidean distance on Shopping AdaBoost | tree ensemble counterfactual search · Shopping AB | d_mean (Euclidean): 0.126 -> 0.028 | Improved |
| Use FOCUS instead of Feature Tweaking on single Decision Trees, Euclidean distance | tree counterfactual search · HELOC DT | d_mean (Euclidean): 0.120 -> 0.133 | Hurt |
| Use FOCUS instead of Feature Tweaking on Decision Trees, Euclidean | tree counterfactual search · Shopping DT | d_mean (Euclidean): 0.119 -> 0.142 | Hurt |
| Use FOCUS instead of Feature Tweaking, Euclidean on Wine Quality Decision Tree | tree counterfactual search · Wine Quality DT | d_mean (Euclidean): 0.269 -> 0.268 | No clear effect |
| Use FOCUS instead of Feature Tweaking, Cosine distance on Wine Quality Decision Tree | tree counterfactual search · Wine Quality DT | d_mean (Cosine): 0.030 -> 0.003 | Improved |
| Use FOCUS instead of Feature Tweaking with Manhattan distance on Random Forests | random forest counterfactual search · Wine Quality RF | d_mean (Manhattan): 0.223 -> 0.312 | Hurt |
| Use FOCUS instead of Feature Tweaking with Manhattan distance on Shopping AdaBoost | AdaBoost counterfactual search · Shopping AB | d_mean (Manhattan): 0.142 -> 0.046 | Improved |
| Compare FOCUS and FT across 36 settings (4 datasets x 3 models x 3 distances) | FOCUS · all datasets | d_mean: 20 settings FOCUS better, 8 FT better, 8 no difference | Mixed |
| Switch from tree ensembles to single Decision Trees (model type) | FOCUS vs FT · all datasets | d_mean | Mixed |
| Use FOCUS instead of DACE with Mahalanobis distance on Wine Quality DT | counterfactual search · Wine Quality DT | d_mean (Mahalanobis): 1.325 -> 0.542 | Improved |
| Use FOCUS instead of DACE with Mahalanobis on Shopping AdaBoost | counterfactual search · Shopping AB | d_mean (Mahalanobis): 3.230 -> 0.303 | Improved |
Closest papers by meaning
- Counterfactual Maps: What They Are and How to Find Them (NeurIPS 2026 · similarity 0.87)
- Robust Stochastic Graph Generator for Counterfactual Explanations (AAAI 2024 · similarity 0.84)
- A General Search-Based Framework for Generating Textual Counterfactual Explanations (AAAI 2024 · similarity 0.84)
- UNR-Explainer: Counterfactual Explanations for Unsupervised Node Representation Learning Models (ICLR 2024 · similarity 0.84)
- Graph Inverse Style Transfer for Counterfactual Explainability (ICML 2025 · similarity 0.83)
- Counterfactual Metarules for Local and Global Recourse (ICML 2024 · similarity 0.83)
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.
