In our current social media age, there is an ever-increasing amount of text data available onthe web. With Twitter being a popular medium for sharing such information, the classification of text data could help in cases of emergencies. In this paper, I explore the workings of various algorithms on classifying disaster tweets. Using the dataset originally created by the company figure-eight, we found that out of Multinomial Naive Bayes, Logistic Regression, Random Forest, BERT, and mixed-BERT-LSTM model, BERT-base-cased model performed best with over 0.82 recall score.