Skip to content

Social Media Pulse

May 25, 2012
tags: ,

Social media stories evolve according to a particular pattern when surprise events are experience by many social media users at once.  An example from Twitter was the earthquake in Oaxaca, MX on 20 March 2012.  The Twitter volume for terms “quake” and “terremoto” are shown in the figure.

The Social Media Pulse for simultaneously observed events can be modeled as a Stationary Poisson Point Process.  The model yields a Gamma distribution,

f_S(t)=\frac{\beta^{-\alpha}(t-t_0)^{\alpha-1}\exp(\frac{-(t-t_0)}{\beta}) }{\Gamma(\alpha)}

and the rate,




(Another, slightly more tractable model analytically for the The Social Media Pulse is to fit with a double exponential, r(t) = r_0(1-\exp(-\alpha (t - t_0))\exp(-\beta (t-t_0)))

Doing this enables consistent comparison of a story across social media publishers as well as comparisons of various stories.

A detailed writeup of the pulse parameters and fitting is available at Social Media Pulse and additional discussion of social media firehoses can be found in my Gnip blog series Taming the Social Media Firehose.

Python code on Github.

No comments yet

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: