Exploring the Potential of Neural Networks for Enhanced Prediction of Societal Impact from Natural Disasters Using LSTM and ELM Model

The availability of data sources has increased dramatically as a result of technological advancements and the free flow of information. The advent of new data sources and significantly larger data sets has presented engineers with a once-in-a-lifetime opportunity to create more accurate and useful models with real-world applications. Predicting and assessing people's health using big data before and after a threat occurs is laid out in this work using a probabilistic technique. It is crucial to follow this exact procedure while performing preprocessing, feature selection, and model training for optimal accuracy. After the proposed approach have filtered out the irrelevant data, the proposed approach can begin to prepare it for content analysis by using preprocessing and data filtering. Commonly known as the Shannon entropy, it is a crucial metric in the field of information theory and is employed in feature selection. To train the model, an LSTM-ELM was utilized. The proposed method outperforms legacy options like LSTM and ELM. The accuracy went up by 91.38% once the method was applied.