Construction Of Disaster Information Collection System Based On Internet News

In recent years,natural disasters have occurred on a global scale,causing serious casualties and property losses.The disaster information is an important basis for emergency decision-making when a disaster occurs.Collecting disaster information in a timely,accurate,and comprehensive manner enables the rescue force to quickly reach the affected location and carry out disaster relief operations.The “National Comprehensive Disaster Prevention and Mitigation Plan(2016-2020)” promulgated by the General Office of the State Council mentioned the need to strengthen basic theoretical research and research and development of key technologies,promote “Internet Plus,” big data,Internet of things,cloud computing,geographic information,The application of new concepts,technologies,and new methods,such as mobile communications,will increase the ability of information acquisition during disasters.This article uses the techniques of machine learning,text categorization,text information mining,and Internet reptiles to conduct research on the collection of disaster information based on Internet news,with a view to expanding the sources of information at the time of disaster,improving the capacity of disaster information collection,and providing information support for emergency management.,And with Hunan Province as a case area,develop system prototypes.The work of this article mainly includes the following aspects:News classifier build.Using machine learning and text categorization methods,news categorizers were constructed from collected news corpora,and the performance differences between different text categorization methods,different text representation methods,balanced data sets and unbalanced data sets,and integrated models and individual models were compared.The final result shows that the word vector model is used as the text representation method.The performance of the integrated classifier composed of the four news classifiers trained under equalization data is the best,with an F1 value of 0.926.This classifier is used as the final News classifier in system.Disaster Information Extraction Research.Through the analysis of the content of the news body,the structural rules of the time information,location information and damage information in the news content are studied.The extraction rules are compiled according to the laws,and the purpose of structured storage of text information is finally achieved.With the aid of the Geode Geocoding API,the location information is converted into latitude and longitude information,which provides a basis for the spatial display of location information.System prototype design and development.On the basis of design and research,the use of Python,HTML,JavaScript,WebGIS and other development technologies to build disaster-oriented information collection system for Internet news,to achieve automatic news collection,automatic classification of news based on news headlines,the news of the main damage Information extraction and structuring,map display of disaster location information,and verification of the system in conjunction with the actual operation of the system to verify the feasibility of the system.