Location based bursty event detection and information dissemination using influencers in Twitter

Abstract

Abstract In the intelligence information sharing era, social media paved a way to share information around the world without any moratorium, which helps more in times of crisis. Influencers are a central node who play an important role in information diffusion throughout the community. Identifying the influencer and predicting the geospatial location of the bursty event during a crisis are the most challenging task in an emergency. The proposed location based bursty event detection and influencer identification helps to send the emergency alert message of disasters to the communities through the influencer at the right time. The traditional machine learning models and Bidirectional Encoder Representations from Transformer model are trained to detect emergency event tweets, in which the Valence Aware Dictionary for Sentiment Reasoning + Count Vector + Ensemble model outperforms well with 98% accuracy. The location of the tweets is extracted by a location tagger and Name Entity Recognizer using Natural Language Processing. The Louvain based Harmonic Centrality Algorithm is used to detect user communities based on the location of users mentioned in the tweets, where the community of the network obtains a modularity of 0.75. The integration of harmonic centrality algorithm identifies the influencers of each community in the network which helps in reducing the disaster risk at the time of emergency as well as early dissemination of information throughout the network community.