Distributed Acquisition And Analysis Methods Of Disaster-Related Public Opinions Based On Spatiotemporal And Theme Features

As a unique source of information during natural disasters,social media contains a wealth of disaster-related and spatiotemporal information,which is of great significance to the disaster situational awareness,Prevention and mitigation.However,the acquisition,storage,processing and analysis of disaster-related public opinions are facing major challenges.Therefore,how to effectively improve the efficiency of data acquisition,achieve efficient storage retrieval,aggregate disaster information from unstructured social media data and explore the spatiotemporal distribution trends to assist timely response and assessment are currently urgent problems to be solved.Based on Sina Weibo,the most popular social media platform in China,this paper builds a distributed acquisition and analysis framework of disaster-related public opinions based on its spatiotemporal and theme features to reflect real disaster events from another perspective.The specific research content are as follows:1)Accurate crawling and efficient storage technology under the distributed framework.In order to solve the problems of lacking topic relevance and selectivity in the existing microblog information collection methods,a crawling strategy that takes into account the characteristics of disaster topics is proposed.Considering the diversified spatiotemporal retrieval requirements,this paper put forward a distributed storage architecture based on QBPSTR-tree.On this basis,through the re-customization of Scrapy framework and the optimization and improvement of key technology modules,the disaster data acquisition and storage framework for Sina Weibo platform is designed and implemented which can provide accurate and reliable data support for situational awareness2)Spatiotemporal-thematic analysis method of disaster-related public opinions.This paper constructs a topic classification model based on topic semantics and spatiotemporal constraints.Fully considered the semantics and spatiotemporal aggregation of disaster data,the model can extract and classify disaster-related information in real time.Considering the influence of the spatiotemporal distribution heterogeneity of social media users,a spatiotemporal weighted model of disaster?related data based on user activity is proposed.on this basis,analyzing spatiotemporal-thematic characteristics of disaster-related public opinions from the perspective of time and space to explore the relationship between disaster?related data and actual disaster events and provide a new analytical perspective for the exploration of the spatiotemporal pattern of disaster events.The experimental platform was built and the "Typhoon Mangkhut" was selected as a typical case to verify the effectiveness of the research method.The research results show that the proposed method can timely and accurately obtain the disaster-related public opinions from the massive microblog data in response to the spatiotemporal theme of disaster events,reduce the interference of non-related data,and improve the storage and retrieval efficiency under large-scale data volume scenarios.Through the topic classification model,combined with spatiotemporal big data analysis technology and dynamic visualization methods,this paper comprehensively aggregates the disaster theme,which can reflect the trend and provide analytical references for the timely response to real disaster events.Furthermore,the weighted data can better reflect the spatial and temporal pattern of disaster events.