Text mining study of microblog activity on heatwaves

Evidence shows that heatwave events are increasing all around the world. They cause massive impact on public health, human constructions, and economy. As a small but progressive step, we performed data mining techniques on the famous microblog, Twitter, to find messages sent about this environmental phenomenon. We focused on heatwave-related tweets, and collected all containing the keyword #heatwave. The collecting process started from September 5th for nine weeks. We applied eight classification algorithms (Bayes Network, Nave Bayes, Multinomial Nave Bayes, Decision Tree, Random Forest, KNN, SVM, Maximum Entropy) to learn the patterns of related tweets and create proper classification models to classify new tweets. We achieved a high f-score (more than 90) in classifications. Our findings confirm that social media reflects the severe heatwave events.