Modularity-based selection of the number of slices in temporal network clustering

Patrik Seiron, Axel Lindegren, Matteo Magnani, Christian Rohner, Tsuyoshi Murata, Petter Holme

Abstract: A popular way to cluster a temporal network is to transform it into a sequence of networks, also called slices, where each slice corresponds to a time interval and contains the vertices and edges existing in that interval. A reason to perform this transformation is that after a network has been sliced, existing algorithms designed to find clusters in multilayer networks can be used. However, to use this approach, we need to know how many slices to generate. This chapter discusses how to select the number of slices when generalized modularity is used to identify the clusters.