Mining Twitter Data on COVID-19 for Sentiment analysis and frequent patterns Discovery

Abstract

A study with a societal objective was carried out on people exchanging on social networks and more particularly on Twitter to observe their feelings on the COVID-19. A dataset of more than 600,000 tweets with hashtags like COVID and coronavirus posted between February 27, 2020 and March 25, 2020 was built. An exploratory treatment of the number of tweets posted by country, by language and other parameters revealed an overview of the apprehension of the pandemic around the world. A sentiment analysis was elaborated on the basis of the tweets posted in English because these constitute the great majority. On the other hand, the FP-Growth algorithm was adapted to the tweets in order to discover the most frequent patterns and its derived association rules, in order to highlight the tweeters insights relatively to COVID-19.