The effect of temporal aggregation level in social network monitoring

Abstract

Social networks have become ubiquitous in modern society, which makes social network monitoring a research area of significant practical importance. Social network data consist of social interactions between pairs of individuals that are temporally aggregated over a certain interval of time, and the level of such temporal aggregation can have substantial impact on social network monitoring. There have been several studies on the effect of temporal aggregation in the process monitoring literature, but no studies on the effect of temporal aggregation in social network monitoring. We use the degree corrected stochastic block model (DCSBM) to simulate social networks and network anomalies and analyze these networks in the context of both count and binary network data. In conjunction with this model, we use the Priebe scan method as the monitoring method. We demonstrate that temporal aggregation at high levels leads to a considerable decrease in the ability to detect an anomaly within a specified time period. Moreover, converting social network communication data from counts to binary indicators can result in a significant loss of information, hindering detection performance. Aggregation at an appropriate level with count data, however, can amplify the anomalous signal generated by network anomalies and improve detection performance. Our results provide both insights on the practical effects of temporal aggregation and a framework for the study of other combinations of network models, surveillance methods, and types of anomalies.

The value of summary statistics for anomaly detection in temporally evolving networks: A performance evaluation study

“…In some cases, performance is better in count settings, but not in others. 13 We make note the design of S * method is to detect if a change occurred rather than to identify the time-points said change occurred in Priebe et al 14 and Zhao et al 12 That is, when monitoring the scan statistic, the method is expected to signal at best only a few times if any anomaly occurred. In AUC results reported in Tables A1-A4 in Appendix A3, the detection ability is mainly best for M − t and M + t methods in DLSM settings while the W t method does detect best in DDCSBM settings.…”

“…Similarly, several papers contain studies of network monitoring under specific parametric modeling frameworks. [10][11][12][13] However, such monitoring methods work under the assumption of a specific network model and cannot be extended to the general task of network monitoring without model assumptions.…”

“…A notable point made by Woodall et al 8 , which was re‐iterated in Sengupta and Woodall, 9 is that even with variation in parameter settings comparable relevant work in the literature are not available for detecting abrupt changes in a stream of time‐varying networks. Similarly, several papers contain studies of network monitoring under specific parametric modeling frameworks 10‐13 . However, such monitoring methods work under the assumption of a specific network model and cannot be extended to the general task of network monitoring without model assumptions.…”

Discussion of “Statistical methods for network surveillance”

“…Zhao et al studied the properties of the scan method of Priebe et al under the assumption of the Erdős and Rényi model. In addition, we have studied monitoring under the assumption of the degree‐corrected stochastic block model (see the works of Wilson et al, Zhao et al, and Yu et al). Zhao et al showed that there can be a substantial loss of information in using unweighted graphs when weighted graphs are available.…”

Rejoinder to “Statistical methods for network surveillance”

“…It has been pointed out that the frequency with which observations are sampled deterministically influences the speed with which anomalies are identified . A practical reality is that the sampling frequency is sometimes fixed by the incumbent data acquisition scheme.…”