A Novel Approach to Handle Disaster with Twitter Reports by Deep Learning

During normal or man-made disasters, individuals utilize web-based social networking websites like Twitter to give an account of the emergency and calamity related data. Networking is vital during seasons of emergency. This paper advocates for the improvement of mechanized emergency by deep learning frameworks, and for the association and status of proficient disaster reactions by catastrophe moderation groups. We achieve this goal by comparing and contrasting the working of several deep learning networks such as LSTM network, GA network, DB networks, CN networks, GRU networks. Parameters such as precision, F1 score, accuracy and recall are compared to find the best formation model. To make compelling and effective catastrophe analysis, the continuous framework can recognize and channel tweets associated with disasters. From this study, CNN has been identified as the best-performing model, with an accuracy of 90%. Following CNN, the performances of LSTM and GRUN are observed to be around 89%. Finally, GAN and DBN yield a performance of 88%. Through the integration of deep learning skills, industry knowledge, and community involvement, resilient and adaptable systems for disaster preparedness, monitoring, and response can be built.