Research map / Learning theory and algorithms / Prediction and decision losses
Under-bagging Nearest Neighbors for Imbalanced Classification
Hanyuan Hang, Yuchao Cai, Hanfang Yang, Zhouchen Lin
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 |
|---|---|---|---|
| add majority under-sampling (B=1) to k-NN | k-NN · APSFailure | AM: 0.7917 -> 0.9401 | Improved |
| add majority under-sampling (B=1) to k-NN | k-NN · CensusIncomeKDD | AM: 0.6799 -> 0.8412 | Improved |
| add majority under-sampling (B=1) to k-NN | k-NN · Bitcoin | AM: 0.5638 -> 0.6630 | Improved |
| add majority under-sampling (B=1) to k-NN | k-NN · OccupancyDetection | AM: 0.9930 -> 0.9941 | No clear effect |
| increase bagging rounds from B=1 to B=5 (ablation) | under-bagging k-NN · p53Mutants | AM: 0.8824 -> 0.9006 | Improved |
| increase bagging rounds from B=1 to B=5 (ablation) | under-bagging k-NN · Bitcoin | AM: 0.6630 -> 0.6740 | Improved |
| increase bagging rounds from B=1 to B=5 (ablation) | under-bagging k-NN · Adult | AM: 0.8020 -> 0.8047 | Improved |
| increase bagging rounds from B=1 to B=5 (ablation) | under-bagging k-NN · Bitcoin | running time (s): 10.21 -> 11.88 | Hurt |
| halve expected sub-sample size (s=0.5Mn(1)) with B=5 (ablation) | under-bagging k-NN · Adult | AM: 0.8047 -> 0.8046 | No clear effect |
| halve expected sub-sample size (s=0.5Mn(1)) with B=5 (ablation) | under-bagging k-NN · p53Mutants | AM: 0.9006 -> 0.8911 | Hurt |
| halve expected sub-sample size (s=0.5Mn(1)) with B=5 (ablation) | under-bagging k-NN · Bitcoin | running time (s): 11.88 -> 8.11 | Improved |
| use under-bagging (B=1) instead of standard k-NN | k-NN · Bitcoin | running time (s): 16885.49 -> 10.21 | Improved |
Closest papers by meaning
- Restoring balance: principled under/oversampling of data for optimal classification (ICML 2024 · similarity 0.86)
- Bagging in overparameterized learning: Risk characterization and risk monotonization (JMLR 2023 · similarity 0.86)
- Imbalanced-learn: A Python Toolbox to Tackle the Curse of Imbalanced Datasets in Machine Learning (JMLR 2017 · similarity 0.85)
- Learning Curves for Noisy Heterogeneous Feature-Subsampled Ridge Ensembles (NeurIPS 2023 · similarity 0.85)
- Learning under Localized Minority Imbalance (NeurIPS 2026 · similarity 0.84)
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.84)
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.
