When Less Is More: Systematic Analysis of Cascade-Based Community Detection

Abstract

Information diffusion, spreading of infectious diseases, and spreading of rumors are fundamental processes occurring in real-life networks. In many practical cases, one can observe when nodes become infected, but the underlying network, over which a contagion or information propagates, is hidden. Inferring properties of the underlying network is important since these properties can be used for constraining infections, forecasting, viral marketing, and so on. Moreover, for many applications, it is sufficient to recover only coarse high-level properties of this network rather than all its edges. This article conducts a systematic and extensive analysis of the following problem: Given only the infection times, find communities of highly interconnected nodes. This task significantly differs from the well-studied community detection problem since we do not observe a graph to be clustered. We carry out a thorough comparison between existing and new approaches on several large datasets and cover methodological challenges specific to this problem. One of the main conclusions is that the most stable performance and the most significant improvement on the current state-of-the-art are achieved by our proposed simple heuristic approaches agnostic to a particular graph structure and epidemic model. We also show that some well-known community detection algorithms can be enhanced by including edge weights based on the cascade data.

Community structure recovery and interaction probability estimation for gossip opinion dynamics

“…As a result, there is a growing interest in community detection for networked systems based on state observations. Maximum likelihood methods applied to cascade data are introduced in [2], [7]. The paper [2] also proposes a two-step procedure: first the underlying network is recovered and then agents are grouped from the network estimates.…”

“…Maximum likelihood methods applied to cascade data are introduced in [2], [7]. The paper [2] also proposes a two-step procedure: first the underlying network is recovered and then agents are grouped from the network estimates. Blind community detection [4], [5], [6] uses sample covariance matrices of agent states to recover the community structure.…”

“…Often only state dynamics evolving over the network are observable, rather than the network itself. Hence, a growing number of studies have been investigating community detection based on state observations [2], [3], [4], [5], [6], [7]. Lacking network information makes community detection difficult, and it is still not clear how to detect communities based on a single trajectory of networked dynamics, which is considered in this paper.…”

Cascade-based Echo Chamber Detection

“…While some studies (e.g. [10]) focus mainly on the former, in this work we exploit, together with the social graph, the polarity of information pieces and their cascades, which have been shown to be more effective in detecting communities [37] and user-level stances [2].…”

Inferring Diffusion Network from Information Cascades using Transitive Influence

“…To study the behavior of people in a social network, the initial requirement is to infer the network structure from the observed data. Inferring the network structure of neurons in neuroscience [13], sentiment in online social networks [14], [15], community detection [16], or the genes in biology [17], [18] are similar points of interest in current researches. The aim of this article is investigating an epidemiology approach to infer the structure of an influence network from a set of information cascades, i.e., the time history of various events occurred in a network.…”