Influence maximization on temporal networks

TL;DR

This paper investigates influence maximization on temporal networks, highlighting the importance of early dynamic information over static or aggregated data for identifying influential spreaders.

Contribution

It introduces a framework for influence maximization on temporal networks and demonstrates the critical role of early dynamic knowledge in selecting effective spreaders.

Findings

Static or aggregated network knowledge is ineffective for influence maximization.

Early-stage dynamic information is crucial for identifying influential spreaders.

Perfect static knowledge does not significantly improve influence maximization.

Abstract

We consider the optimization problem of seeding a spreading process on a temporal network so that the expected size of the resulting outbreak is maximized. We frame the problem for a spreading process following the rules of the susceptible-infected-recovered model with temporal scale equal to the one characterizing the evolution of the network topology. We perform a systematic analysis based on a corpus of 12 real-world temporal networks and quantify the performance of solutions to the influence maximization problem obtained using different level of information about network topology and dynamics. We find that having perfect knowledge of the network topology but in a static and/or aggregated form is not helpful in solving the influence maximization problem effectively. Knowledge, even if partial, of the early stages of the network dynamics appears instead essential for the identification of quasioptimal sets of influential spreaders.