Fingerprinting Concepts in Data Streams with Supervised and Unsupervised Meta-Information
Ben Halstead 1 , Yun Sing Koh 1 , Patricia Riddle 1 , Mykola Pechenizkiy 2 , Albert Bifet 3 , Russel Pears 4 · 2026
Read Fingerprinting Concepts in Data Streams with Supervised and Unsupervised Meta-Information by Ben Halstead 1 , Yun Sing Koh 1 , Patricia Riddle 1 , Mykola Pechenizkiy 2 , Albert Bifet 3 , Russel Pears 4 (2026) 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-2603-11094.txt.
Description
Ben Halstead and coauthors fingerprint the concepts inside data streams. These concepts are the stationary stretches between concept drifts. Systems can then tell whether data matches a known concept or a new one. Earlier representations used only a few meta-information features. They often could not distinguish concepts. This left systems exposed to drift. Their FiCSUM framework builds fingerprints from many supervised and unsupervised features. It uses a dynamic weighting that learns which features signal drift. It beats state-of-the-art methods across eleven datasets.
Catalog record
| CREATOR | Ben Halstead 1 , Yun Sing Koh 1 , Patricia Riddle 1 , Mykola Pechenizkiy 2 , Albert Bifet 3 , Russel Pears 4 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 69182 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.11094