Spatial Analysis Of Internet Sensation Based On Social Meadia

Earthquake is one of the most destructive natural disasters.At present,the existing scientific and technological means are difficult to predict earthquakes accurately.Therefore,comprehensive emergency rescue measures are an important guarantee for reducing post-disaster losses and the accurate assessment of the disaster plays a crucial role.How to obtain disaster information accurately and accurately in real time has always been a hot spot in disaster research.The traditional research mainly extracts disaster information based on remote sensing satellite imagery,field survey and other methods,but with the development of mobile internet and the wide application of location service(LBS),the application enables everyone to share their geographic location information in real time.This phenomenon has brought new entrances and directions to the study of geographical disasters.In the event of an earthquake,the social network and the public opinion data transmitted in the media have very rich disaster information,which can reflect the disaster situation to a certain extent,and provide auxiliary decision-making for the government's macro-control and organization of rescue operations.However,the network information content is complicated and the structure is disordered.How to extract the disaster information in the complex network public opinion information and combine it with its own location information for spatial analysis is of great research significance.Firstly,the paper obtains the microblog sign-in data for the whole year of 2017 through the web crawler technology based on Weibo API,integrates the check-in data and extracts the points of interest,and proposes a user distribution model based on this.At the same time,this paper takes the Jiuzhaigou earthquake as a research case,obtains the microblog data related to the earthquake within one month after the earthquake,and weights the geographical location of the disaster information through the user distribution model to eliminate the spatial heterogeneity caused by the user distribution.The interference is finally analyzed by spatial interpolation of the weighted disaster data to obtain the distribution results of the Jiuzhaigou earthquake disaster.The main research contents and conclusions of this paper are as follows:(1)Microblog data collection and interest point update research combined with simulated landing technology.Based on the Weibo API and the simulation landing method,this paper realized the rapid collection of Weibo data.In the end,the total number of signed data in 2017 was about 23.6 million and the data related to Jiuzhaigou earthquake was 280,000.The experimental results show that the simulated landing technology can effectively solve the user authentication problem of social media,and break the frequency of API calls,which greatly improves the efficiency and quality of data collection.Based on the check-in data,the existing POI data set is updated,and finally about 230,000 points of interest are obtained,and each POI point has an average of 104 check-in information.The experimental results show that the updated interest points are rich in content and complete in information,which proves that social media sign-in data has better applicability in POI update.(2)Microblog short text recognition research.Considering the complexity of the text content,we manually classify the training set by labeling each text.Then we use the ICRCALS system to process the disaster text,and the Word2vec tool is used to extract the feature word vector of the training set.Finally,we train the convolutional neural network through the training set classification result and the feature word vector to obtain the network short text classification model and the model.Meanwhile check the feasibility of this model.The results show that the overall classification accuracy of the model can reach 86.6%,so we use this model to classify the remaining public opinion data.The results show that the classification model filters nearly 55%of meaningless data,leaving 103,559 microblogs.The above results show that the convolutional neural network algorithm has good applicability in the network short text recognition problem,which can effectively solve the dimensionality disaster and local optimal solution of the text vector,and improves the classification efficiency and accuracy of the Internet short text.(3)Construction of user distribution model.Considering that the user's own spatial distribution has extremely high spatial heterogeneity,if the influence of spatial heterogeneity is not eliminated,it will be subjectively considered that the user's sign-in area is the occurrence area or the spread area of the disaster situation,which will disturb the extraction of the disaster.Therefore,based on the updated POI dataset,this paper constructs a user distribution model to calculate the active weight of the checked-in data and perform spatial autocorrelation analysis.Experiment shows that the global Moran's I with active weight has a significant improvement compared with the disaster POI data,and the average neighbor point has a maximum value of 0.398 when it is near 90,which indicates that the user distribution model can effectively improve the spatial autocorrelation of attribute data and reduce The interference caused by spatial heterogeneity.(4)Temporal and spatial variation on network earthquakes and driving forces analysis.This paper studies on the temporal and spatial variation on network earthquakes and driving forces analysis.The experimental results show that the development of the response of the public to the earthquake disaster on the network is in line with the evolution of the public opinion event,and in the sudden and prolonged period of the event.During the period,the public responded to a high degree of aggregation and migration in space.Based on the actual development of the disaster situation,it is found that the aftershock factor,traffic factor and the evacuation factor of the tourists are the main reasons for the above phenomenon.This phenomenon indicates that the temporal and spatial evolution of the network earthquake sentiment has certain relevance to the development of the disaster situation and can be used as a disaster situation.