A Markov model for inferring flows in directed contact networks

TL;DR

This paper introduces an inhomogeneous Markov model for directed contact networks to infer flows, enabling data reduction and anomaly detection in various applications like communication and epidemiology.

Contribution

It presents a novel Markov model tailored for DCNs, facilitating flow inference and anomaly detection in temporal network data.

Findings

Effective data reduction in DCNs

Successful anomaly detection example

Model applicability to various domains

Abstract

Directed contact networks (DCNs) are a particularly flexible and convenient class of temporal networks, useful for modeling and analyzing the transfer of discrete quantities in communications, transportation, epidemiology, etc. Transfers modeled by contacts typically underlie flows that associate multiple contacts based on their spatiotemporal relationships. To infer these flows, we introduce a simple inhomogeneous Markov model associated to a DCN and show how it can be effectively used for data reduction and anomaly detection through an example of kernel-level information transfers within a computer.