Approximating Temporal Networks from Aggregated Network Data

Temporal network data is typically provided in the form of temporal edge lists, where pairs of connected nodes are associated with their interaction timestamps, enabling construction of network snapshots in accumulative or interval manners. For data privacy and/or data storage purposes, data can be instead available only in an aggregated form with the first- and last interaction times and the frequency. In this paper we address the problem of reconstructing temporal networks from aggregated network data. We propose an approach to derive approximate interaction times through linear interpolation over edge time-spans. Our experiments show the feasibility of constructing approximate temporal networks, where the errors are minimal for longer intervals and/or intervals closer to the boundaries of the system lifespan.