Network Clustering Based on Degree Fairness

Abstract

In network data clustering, there is an issue where bias based on node degree affects the clustering results. This bias causes nodes that should have different characteristics to be classified into the same group, making it difficult to discover the true characteristics hidden in the data. In this paper, we define the state where fair results can be obtained without being affected by degree as “degree fairness” and aim to improve it. In this study, we propose a new method that introduces constraints related to node degree into the k-medoids algorithm. To verify the effectiveness of the proposed method, we conduct comparative experiments using conventional methods.