Detecting Periods of Significant Increased Communication Levels for Subgroups of Targeted Individuals

Abstract

This paper investigates detecting significant increases in communication patterns and levels between small groups of individuals within a moderate-size targeted group. Potential applications range from trying to establish emerging thought leaders within an organisation to the detection of the planning stages of a crime. The scan statistic is a popular choice for monitoring and detecting spatio-temporal outbreaks, but it is difficult to apply to large-scale target groups because of the computational effort required. When monitoring communication levels between thousands of people, the number of combinations of people whose communication may have increased is very high, and to scan through all of these to find which combinations have increased communications significantly is an enormous task. A successful surveillance plan will have early communication outbreak detection properties and good diagnostic capabilities for identifying individuals contributing to this outbreak. This paper proposes a new computationally feasible approach for detecting communication outbreaks based on exponentially weighted moving average smoothed communication counts between individuals within the network. We apply a cumulative sum of ordered signal-to-noise (SN) ratios for communication counts to flag significant departures from their respective median values. This plan is demonstrated to be efficient at detecting changes in communication levels for a small part of the network and diagnosing who is involved in the outbreak. Copyright © 2015 John Wiley & Sons, Ltd.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 13 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.