Opinions from tweets as good indicators of leadership and followership status

Abstract

Scores of public opinion about two popular world leaders collected from tweets based on the sentiment they exhibited were classified using two Machine learning techniques (Naïve Bayes and Support vector machines), and four features (Words, unigrams, bigrams and negation) for the classification, we found that the Naïve bayes with unigram features attained a high accuracy of up to 90% therefore indicating that tweets can be used to suggest potential candidates in political election and ways to improve a leaders reputation. © 2006-2015 Asian Research Publishing Network (ARPN.