A Comparative Analysis of Multilayer Network Software

Abstract: As the study of multilayer networks has increased in popularity in the recent decades, so has the number of available multilayer network software. However, few studies have compared this software in terms of scalability performance, overall features provided, and their underlying implementations and crucial design decisions. In this thesis, a selection of four multilayer network libraries: Pymnet, Multinet, multiNetX and muxViz were compared thoroughly. This was done by using citation network data from a publicly available dataset and creating a temporal multilayer network model from it. The libraries executed designed queries on this model for finding suitable operations to test their scalability. These operations were aggregation and degree calculation and the implementation of this whole analysis was written in Python and R. Pymnet and Multinet were applicable in all use cases, but muxViz performed the best when the scalability of all software was tested for the aggregation operation. When testing the scalability of calculating the degree of the nodes in a temporal network, weaknesses in the software were discovered. This work can be extended by testing additional multilayer network software and using several different datasets to construct different multilayer networks that are not restricted to one dimension.