TweetSense: Navigating Disaster Signals Using Machine Learning and Deep Learning Techniques

A disaster is a sudden, catastrophic event that causes significant damage, destruction, and disruption to the normal functioning of society, often resulting in loss of life, property, and livelihoods. Disasters can be natural (earthquakes, floods, wildfires) or human-made (industrial accidents, terrorist attacks, wars). Disasters can have wide-ranging impacts on communities, economies, and the environment, requiring emergency response and recovery efforts to mitigate their effects and restore normalcy. With disasters being too frequent in the modern world, the analysis of tweets cannot be an easy task for manual analysis. To overcome this, an automated model is to be built, which can reduce the manual analysis time for emergency responders. Socialmedia, with Twitter being a prime example, acts as a primary channel for disseminating information regarding disasters. The ability to predict relevant tweets holds immense potential in the development of early warning systems, facilitating timely actions by authorities and emergency responders to mitigate the impact of disasters. The approaches followed to achieve the tweet classification are Bag of Words, Term Frequency-Inverse Document Frequency, Word2Vec, followed by machine learning & deep learning methodologies. The CNN model is notably achieving a higher accuracy of $\mathbf{8 4 . 1 \%}$ compared to other models.