Ashifa, K.M.Babu, JobiSafaei, MehdiArumugam, Thangaraja2024-09-112024-09-1120242520-8195https://doi.org/10.1007/s41976-024-00115-1https://hdl.handle.net/11363/8538Many nations have created their own frameworks for disaster risk management (DRM) in response to the rising frequency of catastrophes that cause significant losses. Finding shelter is one of the most pressing demands of anyone impacted by a disaster. While the abundance of catastrophe data is already assisting in the saving of lives, it is necessary to quickly combine a broad variety of data in order to detect building damages, determine the need for shelter, and choose the best locations to set up emergency shelters or settlements. This research suggests a machine learning (ML) approach that seeks to fuse as well as quickly evaluate multimodal data in order to fill this gap and advance complete evaluations. This study suggests a unique approach to managing environmental disasters that is based on the analysis of geographical data using a machine learning model. Here, the input is a geospatial picture of a region that frequently experiences disasters, which is then smoothed and noise-removed. Then, a fuzzy clustering–based deep spatial reinforcement model (FCDSR) was used to choose the characteristics of the processed data. Multimodal Dirichlet allocation–based LSTM (long short-term memory) logistic correspondence algorithm (MDALLCA) was used to extract the chosen features. For various catastrophe datasets, experimental analysis is done in terms of prediction accuracy, precision, F-measure, and ROC. Our findings indicate possible locations with a high density of impacted people as well as infrastructure damage during the course of the crisis. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2024.eninfo:eu-repo/semantics/closedAccessEnvironmental disaster management; Fuzzy clustering; LSTM correspondence algorithm; Machine learning; Spatial data analysisArticle10.1007/s41976-024-00115-12-s2.0-85201539983N/A