Research map / Label-efficient and robust learning / Recognition and long-tail
Classification of Imbalanced Data with a Geometric Digraph Family
Artür Manukyan, Elvan Ceyhan
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 CCCD classifiers instead of k-NN when one class support is embedded in the other | CCCD classifiers · FX=U(0,1)^d, FY=U(0.3,0.7)^d, equal class sizes | AUC: nearly 0.20 AUC more than k-NN with increasing dimensionality | Improved |
| SVM instead of P-CCCD and RW-CCCD in embedded-support setting | classifier choice · FX=U(0,1)^d, FY=U(0.3,0.7)^d, equal sizes | AUC: SVM about 0.05 more AUC than P-CCCD and RW-CCCD | Hurt |
| rebalance class sizes locally (m=50, n=200 or 1000) so overlap region is less imbalanced (ablation) | k-NN and C4.5 vs CCCD · FX=U(0,1)^d, FY=U(0.3,0.7)^d | AUC: k-NN 0.05 more AUC than P-CCCD; RW-CCCD 0.05 more than k-NN | Mixed |
| increase local class imbalance in overlap region (ablation) | k-NN and C4.5 · overlapping uniform boxes | AUC | Hurt |
| use RW-CCCD in imbalanced and moderately overlapping data | RW-CCCD vs non-CCCD classifiers · U(0,1)^d vs U(delta,1+delta)^d | AUC: at least 0.05 more AUC | Improved |
| use CCCD when classes are balanced or barely overlap | CCCD classifiers · U(0,1)^d vs U(delta,1+delta)^d | AUC | No clear effect |
| use C4.5 instead of CCCD with well separated supports | classifier choice · disjoint uniform boxes | AUC: C4.5 over 0.04 more AUC than CCCD | Hurt |
| use P-CCCD instead of RW-CCCD when classes are well separated (ablation) | CCCD variants · disjoint uniform boxes | AUC | Improved |
| use RW-CCCD with classes imbalanced and slightly far apart (ablation) | RW-CCCD · disjoint uniform boxes | AUC | Hurt |
| use unpruned C4.5 instead of pruned (ablation) | C4.5 · all simulation settings | AUC | Improved |
| use RW-CCCD vs strong imbalance-robust classifiers | RW-CCCD · overlapping imbalanced simulation | AUC: 0.05 more than C5.0 and ESVM; 0.05 less than others at low delta and q | Mixed |
| use RW-CCCD prototypes vs SVM support vectors | RW-CCCD · simulation, training set reduction | model size | Improved |
Closest papers by meaning
- Class-Imbalanced Graph Learning without Class Rebalancing (ICML 2024 · similarity 0.86)
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.86)
- Cluster-guided Contrastive Class-imbalanced Graph Classification (AAAI 2025 · similarity 0.85)
- Balanced Data, Imbalanced Spectra: Unveiling Class Disparities with Spectral Imbalance (ICML 2024 · similarity 0.84)
- CLIMB: Class-imbalanced Learning Benchmark on Tabular Data (NeurIPS 2025 · similarity 0.84)
- Geometric Imbalance in Semi-Supervised Node Classification (NeurIPS 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.
