Neighbor displacement-based enhanced synthetic oversampling for multiclass imbalanced data
I Made Putrama, Peter Martinek · 2025
Read Neighbor displacement-based enhanced synthetic oversampling for multiclass imbalanced data by I Made Putrama, Peter Martinek (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-04099.txt.
Description
I Made Putrama and Peter Martinek tackle lopsided training data, where some classes have far fewer examples than others. Rare classes often matter most, yet models learn them poorly. Making fake extra examples helps, but noisy or poorly placed fakes can blur the true patterns and hurt results.
Their method, NDESO, first nudges noisy points toward the center of their own group, based on how far they sit from neighbors. It then adds random copies to balance the classes. The team tested the method against fourteen rivals on nine learning models, using both made-up and twenty real data sets. NDESO earned the best average scores and the best overall rank. The authors offer it as a practical fix for skewed data.
Catalog record
| CREATOR | I Made Putrama, Peter Martinek |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-04099