Microblog Disaster Information Aggregation Based On LDA

The increasing popularity of social media such as Microblog has provided new ideas for disaster response.The disaster emergency task cannot be carried out if there is no data.Microblog builds a bridge for information communication outside the disaster area with its own advantages.So it has become one of the important sources of disaster information timely acquisition in disaster events.However,the unstructured and fast-updated nature of Microblog has made it difficult to make full use of this information.Therefore,how to quickly aggregate more comprehensive disaster information from unstructured Microblog data to assist disaster emergency decision-making is a problem that needs to be solved urgently.For the current situation,where microblog information aggregation applications in the field of disasters are less and the aggregated information based on these methods is limited and not overall,and surrounding the key problems of aggregation granularity and content richness,this paper proposes a microblog disaster information aggregation method based on LDA.It mainly includes the following four aspects of research:(1)Study of microblog data preprocessing method.The existing text preprocessing method does not consider the characteristics of disasters,which leads to that the results can not meet the aggregation requirements of this paper.This paper is on the base of the existing methods,combined with the disaster domain knowledge,developed a suitable microblog data preprocessing program;(2)Study of disaster thematic model construction method based on LDA.This paper summarizes the three principles of the model parameter estimation method based on the demand for rapid and full-scale aggregation of microblog disaster information,namely,high recognition,large differences between topics,low time complexity,and Gibbs is selected through experimental analysis as a parameter estimation method for topic model;(3)Study of microblog disaster information aggregation method.This paper presents the microblog disaster information aggregation process,focus on the aggregation granularity,do the research on microblog aggregation from data to information,which gradually aggregates disaster information on both time-space distribution information and thematic feature information.Finally,the method of aggregated information visualization is studied so that information can be visually displayed.(4)Design and development of information aggregation platform.On the base of the research in information aggregation platform,this paper has put it into practice through programming languages,and take Super Typhoon 'Hato' that occurred on August 20th 2017 as the example to verify the feasibility of this method.The research results show that the proposed method,aggregates microblog data from massive microblog through disaster keywords and time,which reduces the interference of non-related data and provides an effective data for information aggregation.Through the identification of geographical names of text and tf-idf,the obtained provincial level space distribution information effectively reflects the disaster space distribution and affected degree,and verifies the feasibility of spatial distribution of text data mining;and the disaster thematic feature information through the theme river map,visually shows the development trend of various effective topics during the disaster.This paper more comprehensively aggregates effective disaster information in Weibo and provides great decision support for disaster response work.However,the method of this paper still has some deficiencies and needs to be further explored in the future.