A Multi-Model Approach for Disaster-Related Tweets

Abstract

This research centers around utilizing Natural Language Processing (NLP) techniques to analyze disaster-related tweets. The rising impact of global temperature shifts, leading to irregular weather patterns and increased water levels, has amplified the susceptibility to natural disasters. NLP offers a method for quickly identifying tweets about disasters, extracting crucial information, and identifying the types, locations, intensities, and effects of each type of disaster. This study uses a range of machine learning and neural network models and does a thorough comparison analysis to determine the best effective method for catastrophe recognition. Three well-known techniques, in-cluding the Multinomial Naive Bayes Classifier, the Passive Aggressive Classi-fier, and BERT (Bidirectional Encoder Representations from Transformers) were carefully examined with the ultimate goal of discovering the best strategy for correctly recognising disasters within the context of tweets. Among the three models, BERT achieved the highest performance in analyzing disaster-related tweets with an accuracy of 94.75%.