Springe direkt zu Inhalt

Drift-Aware Multi-Memory Model for Imbalanced Data Streams

Prof. Dr. Eirini Ntoutsi, Amir Abolfazli – 2021

Online class imbalance learning deals with data streams that are affected by both concept drift and class imbalance. Online learning tries to find a trade-off between exploiting previously learned information and incorporating new information into the model. This requires both the incremental update of the model and the ability to unlearn outdated information. The improper use of unlearning, however, can lead to the retroactive interference problem, a phenomenon that occurs when newly learned information interferes with the old Information and impedes the recall of previously learned information. The problem becomes more severe when the classes are not equally represented, resulting in the removal of minority information from the model. In this work, we propose the Drift-Aware Multi-Memory Model (DAM3), which addresses the class imbalance problem in online learning for memory-based models, namely KNNs. DAM3 mitigates class imbalance by incorporating an imbalance-sensitive drift detector, preserving a balanced representation of classes in the model, and resolving retroactive interference using a working memory that prevents the forgetting of old information. We show through experiments on real-world and synthetic datasets that the proposed method mitigates class imbalance and outperforms the state-of-the-art methods.

Titel
Drift-Aware Multi-Memory Model for Imbalanced Data Streams
Verfasser
Prof. Dr. Eirini Ntoutsi, Amir Abolfazli
Schlagwörter
online learning, class imbalance, concept drift, retroactive interference, multi-memory model
Datum
2021-03-19
Kennung
DOI: 10.1109/BigData50022.2020.9378101
Erschienen in
Proceedings of the 2020 IEEE International Conference on Big Data
Zitierweise
A. Abolfazli and E. Ntoutsi, "Drift-Aware Multi-Memory Model for Imbalanced Data Streams," 2020 IEEE International Conference on Big Data (Big Data), Atlanta, GA, USA, 2020, pp. 878-885, doi: 10.1109/BigData50022.2020.9378101.
Größe oder Länge
8 pages
Rechte
Copyright by IEEE. When citing this work, cite the IEEE-link.
BibTeX Code
@inproceedings{Abolfazli2020drift, title={Drift-Aware Multi-Memory Model for Imbalanced Data Streams}, author={Abolfazli, Amir and Ntoutsi, Eirini}, booktitle={2020 IEEE International Conference on Big Data (Big Data)}, pages={--}, year={2020}, organization={IEEE} }