Jueves, 29 de Septiembre de 2016
12:00 - 13:00
Santiago CESS, Concha y Toro 32C, Santiago
Characterizing Social Network Users for Election Forecast
We present a methodology to identify active, influential, and political Twitter users, and extract their public information from different social networks, in order to determine what kind of people use Twitter to make known their political interest. This can be highly relevant to interpret results of analysis and predictions regarding this social network, such as determining the approval ratings of candidates before the elections. Our methodology considers two kind of different analysis. Firstly, we want to obtain the socio-demographic information of both the most active and the most influential Twitter users in the corpus. Since the personal information provided by Twitter is poor, we need to collect the most personal information as we can, from other social networks, like Facebook. Secondly, we define a novel measure based on the timeline of activity of each user, that allows to find those users that were more active in days with political milestones. We conjecture that users more active in that days are probably more interested in politics. We apply our methodology for an exhaustive analysis of a large corpus of tweets collected during the four-month period approaching the 2015 UK election.