Monitoring user-system interactions through graph-based intrinsic dynamics analysis

Abstract

Abstract-Monitoring the evolution of user-system interactions is of high importance for complex systems and for information systems in particular, especially to raise alerts automatically when abnormal behaviors occur. However current methods fail at capturing the intrinsic dynamics of the system, and focus on evolution due to exogenous factors like day-night patterns. In order to capture the intrinsic dynamics of user-system interactions, we propose an innovative graph-based approach relying on a novel concept of time. We apply our method on two large real-world systems (the Github.com social network and the eDonkey peer-topeer system) to automatically detect statistically significant events in a real-time fashion. We finally validate our results with the successful interpretation of the detected events.

The Expression of Cultural Identities in Hong Kong’s Anti-Extradition Law Amendment Bill Movement: A Semantic Network Analysis of Tweets

“…Modularity measures the strength of the division of a network into clusters and evaluates the divisions the algorithm generated (Newman & Girvan, 2004). Then, we applied the Louvain optimization modularity algorithm (i.e., a tool to calculate modularity properties embedded in Gephi) to detect automatically the nonoverlapping and optimized partitions with the highest modularity score, grouping each word node into a specific cluster and marking with colors (Heymann & Le Grand, 2013). The prominence of a cluster is determined by the number of nodes it contains; thus, the larger the number of nodes in a cluster, the more prominent the cluster.…”

On the use of intrinsic time scale for dynamic community detection and visualization in social networks

“…Whereas extrinsic time is broadly used without notice, previous works have shown that the choice between extrinsic and intrinsic time has a very significant impact both on the measurement of statistical properties of temporal networks and on observation of diffusion processes [32], [33]. Previous results on social networks suggest that intrinsic time is better at characterizing the endogenous dynamics of the network, because extrinsic time is more likely to capture exogenous patterns like day-night activity of users.…”

A matter of time - intrinsic or extrinsic - for diffusion in evolving complex networks

“…Whereas the extrinsic time is broadly used without notice, we have found out in [10] that these two different concepts of time have a high impact on the measurement of statistical properties of temporal networks. Our previous results seem to suggest that intrinsic time is better at characterizing the endogenous dynamics of the network, because extrinsic time is more likely to capture exogenous patterns like day-night activity of users in information networks.…”

“…In this paper, we propose a simple yet innovative approach to study the impact of graph dynamics on diffusion. This methodology does not require any new computation once the diffusion process has been measured; instead of observing the diffusion phenomenon as a function of usual time -e.g., measured in seconds-which we call here extrinsic time, we propose to observe it as a function of what we call intrinsic time [10]. Indeed, this time is intrinsically related to graph dynamics as an intrinsic time slot is not absolute: it corresponds to the appearance of new links in the network, as explained in the following of the paper.…”