IDTCKS: An Intelligent Integrative Approach for Disaster Tweet Classification and Disaster-Related Document Classification Using Knowledge-Driven Hybrid Semantics

There is an immediate requirement for a strategic semantically inclined, semantic web accommodating framework for disaster tweet classification and disaster document and newsfeed documents recommendation to perform disaster risk mitigation and management. This framework proposes an IDTCKS framework for disaster management as a domain of choice, wherein an agent-driven model for integrating auxiliary knowledge from several news and Twitter APIs is encompassed. Subsequently, the TF-IDF discovers informative terms from the documents. The frameworks encompass two distinct classifiers, namely the AdaBoost Classifier to classify the disaster tweets using the upper ontology for the disaster domain as the features and upper ontology is considered by using the LDA to the enriched disaster terms. Subsequently, the upper ontology and the LDA-enriched disaster terms are also used as a feature-driven logistic regression classifier that helps in classifying the dataset. Semantic relevance is achieved using Twitter Semantic Similarity and Horn’s Index with a threshold and step deviance measure and overall precision of 97.39, an accuracy of 98.24, and a very low False Negative Rate of 0.01 has been achieved by the IDTCKS model for disaster classification while an overall precision of 96.57, accuracy of 97.65, and a very low False Negative Rate of 0.02 has been achieved by the IDTCKS model for document classification, which performs much better than other baseline models.