The Pulse of News in Social Media Forecasting Popularity

The Pulse of News in Social Media: Forecasting Popularity
Roja Bandari
?
Sitaram Asur
?
Bernardo Huberman
?
Abstract
News articles are extremely time sensitive by nature. There
is also intense competition among news items to propagate
as widely as possible. Hence, the task of predicting the pop-
ularity of news items on the social web is both interesting
and challenging. Prior research has dealt with predicting
eventual online popularity based on early popularity. It is
most desirable, however, to predict the popularity of items
prior to their release, fostering the possibility of appropriate
decision making to modify an article and the manner of its
publication. In this paper, we construct a multi-dimensional
feature space derived from properties of an article and eval-
uate the e?cacy of these features to serve as predictors of
online popularity. We examine both regression and classi?-
cation algorithms and demonstrate that despite randomness
in human behavior, it is possible to predict ranges of pop-
ularity on twitter with an overall 84% accuracy. Our study
also serves to illustrate the di?erences between traditionally
prominent sources and those immensely popular on the so-
cial web.
1 Introduction
News articles are very dynamic due to their relation to
continuously developing events that typically have short
lifespans. For a news article to be popular, it is essential
for it to propagate to a large number of readers within
a short time. Hence there exists a competition among
di?erent sources to generate content which is relevant
to a large subset of the population and becomes virally
popular.
Traditionally, news reporting and broadcasting has
been costly, which meant that large news agencies dom-
inated the competition. But the ease and low cost of on-
line content creation and sharing has recently changed
the traditional rules of competition for public attention.
News sources now concentrate a large portion of their
attention on online mediums where they can dissemi-
nate their news e?ectively and to a large population. It
is therefore common for almost all major news sources to
have active accounts in social media services like Twitter
to take advantage of the enormous reach these services
?
UCLA.
?
HP Labs.
?
HP Labs.
provide.
Due to the time-sensitive aspect and the intense
competition for attention, accurately estimating the
extent to which a news article will spread on the web
is extremely valuable to journalists, content providers,
advertisers, and news recommendation systems. This
is also important for activists and politicians who are
using the web increasingly more to in?uence public
opinion.
However, predicting online popularity of news arti-
cles is a challenging task. First, context outside the web
is often not readily accessible and elements such as local
and geographical conditions and various circumstances
that a?ect the population make this prediction di?cult.
Furthermore, network properties such as the structure
of social networks that are propagating the news, in?u-
ence variations among members, and interplay between
di?erent sections of the web add other layers of com-
plexity to this problem. Most signi?cantly, intuition
suggests that the content of an article must play a cru-
cial role in its popularity. Content that resonates with
a majority of the readers such as a major world-wide
event can be expected to garner wide attention while
speci?c content relevant only to a few may not be as
successful.
Given the complexity of the problem due to the
above mentioned factors, a growing number of recent
studies [1], [2], [3], [4], [5] make use of early measure-
ments of an item’s popularity to predict its future suc-
cess. In the present work we investigate a more di?cult
problem, which is prediction of social popularity with-
out using early popularity measurements, by instead
solely considering features of a news article prior to its
publication. We focus this work on observable features
in the content of an article as well as its source of publi-
cation. Our goal is to discover if any predictors relevant
only to the content exist and if it is possible to make a
reasonable forecast of the spread of an article based on
content featu res.
The news data for our study was collected from
Feedzilla
1
–a news feed aggregator– and measurements
of the spread are performed on Twitter
2
, an immensely
1
http://wendang.chazidian.com
2
http://wendang.chazidian.com