And the winner is …: Bayesian Twitter-based prediction on 2016 U.S. presidential election
Elvyna Tunggawan, Yustinus Eko Soelistio · 2016
This paper describes a Naive Bayesian predictive model for 2016 U.S. Presidential Election based on Twitter data. We use 33,708 tweets gathered since December 16, 2015 until February 29, 2016. We propose a simple way for data preprocessing which can still achieve 95.8% accuracy on predicting sentiments. The predicted sentiments are used to forecast the U.S. Republican and Democratic parties candidacies. The forecast is compared to the poll collected from RealClearPolitics.com with 26.7% accuracy. However, the true forecasting capacity of the method still have to be observed after the election process come to conclusion.