Identify-Disaster-Tweets

Identifying the tweets that are sent after the disaster occurs are classified to detect the urgent situations and find the people who need help most. The dataset from the website figure eight under the title “Disasters on social media”, is used and there are a total of 10878 samples in this data. By using SVM, neural network, Naive Bayes, and decision trees and using algorithms like bag-of-words (BOW) and part-of-speech (POS) and Term Frequency (TF) - Inverse Document Frequency (IDF) and trying different ngrams, the data we have is split to train, validate and test. According to our results, the BOW algorithm is better than the POS algorithm. However, the neural network performed the best.