Balanced Fair K-Means Clustering
R. Pan, Caiming Zhong, Jiangbo Qian
Abstract
Fairness in clustering has recently received significant attention. The goal of fair clustering is to ensure that a clustering algorithm mitigates or even eliminates bias in the original dataset. Many existing fair clustering algorithms will sometimes generate numerous small clusters to satisfy the fairness constraint. In this article, we present a balanced fair K-means clustering algorithm that integrates a fairness constraint and a balance constraint into the K-means objective function. The proposed model is a tradeoff between the K-means objective and the fairness constraint and their relative importance can be controlled. The balance constraint prevents the generation of small clusters. Experimental results on both real-world and synthetic datasets demonstrate that the proposed method achieves a better fairness performance than some other fair clustering methods, with an acceptable loss of clustering quality in some cases and an improvement in others.