Abstract
Election prediction has been studied in the recent years. However, the previous works focus on counting the number of tweets mentioning candidates to predict the election result. Many reasons cause candidates to win or lose in an election, such as political opinions, social issues, scandals and other reasons. In this paper, we consider a novel viewpoint to predict election results. For a candidate, if the following event sequence happened, “(big event, positive) → (small event, negative) → (big event, positive)”, this candidate will win the election. We consider four approaches to generate the above sequences and then apply the rule-based classifier for predict the election results. A series of experiments are performed to evaluate our approaches and the experiment results reveal that the accuracy of our approaches on predicting election results is over 80% in most of the cases.