Data mining of social manifestations in Twitter: An ETL approach focused on sentiment analysis

Abstract

The objective of this study was to analyze sentiments of users of online social network twitter to understand how people manifested toward the article published by the magazine Veja on 04-18-16 entitled "bela, recatada e do lar" (beautiful, demure and from home) in an attempt to understand how this behavior evolved in two weeks and to assess which events had aroused greater reaction from people. To this end, a data mining technique known as sentiment analysis was used with the help of the ETL (Extract, Transform and Load) methodology and the Naive Bayes probabilistic learning algorithm. Moreover, the null hypothesis was formulated and tested to see whether two events that took place during the collection period influenced, in fact, the polarity of analyzed sentiments in the generated database.