# Class Distribution Monitoring for Concept Drift Detection

arXiv: https://arxiv.org/abs/2210.08470 · HTML: https://mlsanity.com/paper/2210.08470/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 33 references that are in the map.
- First posted on arXiv 16 Oct 2022.

## Builds on (most cited first, 7 of 7)

- [Learning under Concept Drift: A Review](https://mlsanity.com/paper/2004.05785/) (2020)
- [Recent Advances in Open Set Recognition: A Survey](https://mlsanity.com/paper/1811.08581/) (2018)
- [Challenges in Benchmarking Stream Learning Algorithms with Real-world Data](https://mlsanity.com/paper/2005.00113/) (2020)
- [Scan B-Statistic for Kernel Change-Point Detection](https://mlsanity.com/paper/1507.01279/) (2015)
- [NEWMA: a new method for scalable model-free online change-point detection](https://mlsanity.com/paper/1805.08061/) (2018)
- [Concept Drift Detection from Multi-Class Imbalanced Data Streams](https://mlsanity.com/paper/2104.10228/) (2021)
- [Change Detection in Multivariate Datastreams: Likelihood and Detectability Loss](https://mlsanity.com/paper/1510.04850/) (2015)

## Abstract (arXiv)

We introduce Class Distribution Monitoring (CDM), an effective concept-drift detection scheme that monitors the class-conditional distributions of a datastream. In particular, our solution leverages multiple instances of an online and nonparametric change-detection algorithm based on QuantTree. CDM reports a concept drift after detecting a distribution change in any class, thus identifying which classes are affected by the concept drift. This can be precious information for diagnostics and adaptation. Our experiments on synthetic and real-world datastreams show that when the concept drift affects a few classes, CDM outperforms algorithms monitoring the overall data distribution, while achieving similar detection delays when the drift affects all the classes. Moreover, CDM outperforms comparable approaches that monitor the classification error, particularly when the change is not very apparent. Finally, we demonstrate that CDM inherits the properties of the underlying change detector, yielding an effective control over the expected time before a false alarm, or Average Run Length (ARL0).