On the tweet arrival process at Twitter: analysis and applications Articles uri icon

publication date

  • February 2014

start page

  • 273

end page

  • 282

issue

  • 2

volume

  • 25

International Standard Serial Number (ISSN)

  • 2161-5748

Electronic International Standard Serial Number (EISSN)

  • 2161-3915

abstract

  • This work provides a novel measurement-based analysis of the tweet arrival traffic process at Twitter. The analysis considers more than one million total tweets collected at 48 different times of the day (o'clock and half-past every hour). We observe a 3.5-tweet/ms average rate with a valley of 2.5 tweets/ms at 10 AM (GMT+1) and a peak at 3 PM (GMT+1) of about 5 tweets/ms. We further model the traffic pattern as a Gaussian process, and we validate such an assumption with multiple normality tests. Finally, we overview a number of applications where such a model may show its utility, namely infrastructure dimensioning and upgrading, the detection of outlier events, energy efficiency and so on.

subjects

  • Telecommunications

keywords

  • arrival process; average rate; gaussian processes; measurement-based; normality tests; traffic pattern; energy efficiency; social networking (online)