Interval Signature: Persistence and Distinctiveness of Inter-event Time Distributions in Online Human Behavior

Abstract

Interval patterns, or inter-event time distributions that occur in human activity, have long been an interest of many researchers studying human dynamics. While previous studies have mostly focused on characterizing the aggregated inter-arrival patterns or finding universal patterns across all individuals, we focus on the diversity among the patterns of different individuals; the goal of this paper is to understand how persistent an individual's interval pattern is and how distinctive it is from those of the others. We use Wikipedia, me2DAY, Twitter, and Enron email data to study the interval patterns of online human behavior. Our analysis reveals that individuals have robust and unique interval signatures. The interval pattern of a user tends to persist over years, even after coming back from a long hiatus of inactivity, despite considerable change in circadian rhythms. Furthermore, the interval patterns of individuals are highly distinct from that of others. We put our new findings in practical use of identifying users.