The recently-concluded United States Presidential election was the end to one of the most divisive and vitriolic campaign seasons in recent memory. With one of the candidates being active on Twitter, there was a good opportunity to study whether or not the election results could be predicted from patterns in Twitter’s network of users. Percolation theory, statistical physics of complex networks, natural language processing, and machine learning are combined in an analysis of a network of election-related interactions between Twitter users created from retweets, replies, quotes, and hashtags. The analytics pioneered by Alexandre Bovet, Flaviano Morone, and Hernán Makse were able to anticipate the New York Times polling average by up to 13 days, providing a more instantaneous snapshot of general public opinion than classic polling techniques were able to. See the paper by Alexandre Bovet, Flaviano Morone, and Hernán Makse here: Predicting election trends with Twitter: Hillary Clinton versus Donald Trump.
