Statistical Analysis and Classification of Calamity Related Tweets

Abstract

Twitter and other social media platforms have emerged as popular channels for communication during emergencies. Social networks generate huge amounts of data due to the behaviour of their users. A wide range of topics is discussed on social networks, including politics, health issues, and natural disasters. Therefore, public data provides a wealth of information on many topics. Traditional methods of communication have been enhanced in many ways by the Internet. In disaster assessment, Machine Learning (ML) and Artificial Intelligence (AI) algorithms are becoming increasingly popular. The use of micro blogging platforms such as Twitter during natural catastrophes and emergencies generates an increasing number of posts on these platforms. In this paper, we examine natural disasters, including avalanches, tornadoes, hurricanes, droughts, earthquakes, landslides, tsunamis, floods, volcanoes, and wildfires. We extract data from Twitter Network and classify them as disaster and non-disaster tweets as target and non-targets using SVM, Word2Vec, TFIDF, and BERT model.