Springe direkt zu Inhalt

FABBOO - Online Fairness-aware Learning under Class Imbalance

Vasileios Iosifidis, Eirini Ntoutsi – 2020

Data-driven algorithms are employed in many applications, in which data become available in a sequential order, forcing the update of the model with new instances. In such dynamic environments, in which the underlying data distributions might evolve with time, fairness-aware learning cannot be considered as a one-off requirement, but rather it should comprise a continual requirement over the stream. Recent fairness-aware stream classifiers ignore the problem of class distribution skewness. As a result, such methods mitigate discrimination by "rejecting" minority instances at large due to their inability to effectively learn all classes. In this work, we propose FABBOO, an online fairness-aware approach that maintains a valid and fair classifier over a stream. FABBOO is an online boosting approach that changes the training distribution in an online Fashion based on both stream imbalance and discriminatory behavior of the model evaluated over the historical stream. Our experiments show that such long-termin consideration of class-imbalance and fairness are beneficial for maintaining models that exhibit good predictive- and fairness-related performance.

FABBOO - Online Fairness-aware Learning under Class Imbalance
Vasileios Iosifidis, Eirini Ntoutsi
Springer Link
data streams, fairness-aware classification, class-imbalance
DOI: 10.1007/978-3-030-61527-7_11
Erschienen in
Lecture Notes in Computer Science book series (LNCS, volume 12323) . Proceedings of the 23rd International Conference, DS 2020, Thessaloniki, Greece, October 2020.
Iosifidis V., Ntoutsi E. (2020) FABBOO - Online Fairness-Aware Learning Under Class Imbalance. In: Appice A., Tsoumakas G., Manolopoulos Y., Matwin S. (eds) Discovery Science. DS 2020. Lecture Notes in Computer Science, vol 12323. Springer, Cham. https://doi.org/10.1007/978-3-030-61527-7_11.
open access