Sentiment Analysis of Twitter Data Through Machine Learning Techniques

Abstract

Cloud computingCloud computing is a revolutionary technologySentiment analysis for businessesTwitter, governments, and citizens. Some examples of Software-as-a-Services (SaaS) of cloud computing are banking apps, e-mail, blog, online news, and social networksSocial networks. In this chapter, we analyze dataData sets generated by trending topics on TwitterTwitter that emerged from Mexican citizens that interacted during the earthquake of September 19, 2017, using sentiment analysisSentiment analysis and supervised learning, based on the Ekman'sEkman's six emotional modelEkman's model. We built three classifiers to determine the emotionsEmotions of tweets that belong to the same topic. The classifiers with the best accuracy for predicting emotions were Naive BayesNaive Bayes and support vector machineSupport Vector Machine. We found that the most frequent predicted emotions were happinessHappiness, angerAnger, and sadnessSadness; also, that 6.5% of predicted tweets were irrelevant. We provide some recommendations about the use of machine learning techniques in sentiment analysisSentiment analysis. Our contribution is the expansion of the emotions range, from three (negative, neutral, positive) to six in order to provide more elements to understand how users interact with social mediaSocial media platforms. Future research will include validation of the method with different data setsDatasets and emotions, and the addition of new artificial intelligence techniques to improve accuracy.