Geography Reference
In-Depth Information
consumers by Mitchell and Page (
2013
), the variables of age and education are
particularly considered. The study area is greater London, which is known to be a
hotspot of Twitter users; a recent report showed that London is ranked third of the
20 cities world-wide with the largest number of posted tweets (Mediabistro
2012
).
The specific objective is to assess whether the distribution of tweets provides
insights into the demographic group of Twitter users for greater London.
The employed methodology is based on techniques of exploratory spatial data
analysis. The geolocated tweets of a period of 3 months (July to September 2013)
are first separated into day-time and night-time tweets. This separation reveals
differences in the patterns of tweet distribution across London
'
s wards during day
and night. The night-time tweets are then used for the subsequent steps of the
analyses, because of the assumption that residents tweet from their home location
rather at night than during the day.
A second step of pre-processing is the elimination of tweets from Twitter users
with less than three tweets in the data covering 3 months. This elimination is
thought to increase the chance of working with tweets from locals rather than
tourists, although it is certainly an approximation to the issue only.
The analysis then focuses on identifying tweet hotspots. Tweet hotspots are
locations with large numbers of tweets in comparison to population figures. The
hotspots are then analysed regarding their socio-demographic similarities using the
Exploratory Spatial Data Analysis tool GeoDA. Visual exploration supports the
selection of candidate variables for an Ordinary Least Squares (OLS) regression
analysis.
The visual exploration of London
s wards reveals deviations in demographic
characteristics for hotspots outside the city centre. These deviations can be con-
firmed in the statistical analysis, which does not indicate a strong relationship
between tweet counts and socio-demographic characteristics of the population.
Despite the fact that issues like tweeting tourists and lacking details in the demo-
graphic variables may influence these results, the analysis cannot confirm that tweet
numbers can be explained by socio-demographic variables for London.
'
Twitter Data Analyses
Twitter messages are evaluated for various purposes based on their content and/or
location. The objectives are either to understand the behaviour of Twitter users in
detail or draw conclusions about reported phenomena. There are studies that
examine the content of the Twitter messages to describe urban areas (Wakamiya
et al.
2011
), to assess the reaction of the community to physical events such as
earthquakes (Crooks et al.
2013
) or to perform sentiment analysis (Pak and
Paroubek
2010
; Kouloumpis et al.
2011
). Another specific question frequently
investigated is the influence of Twitter on society. For example, Huberman
et al. (
2008
) examined social ties within the Twitter network. Their work showed
that only a diminutive number of people among the friend and follower network is
