Research On Urban Spatio-temporal Dirtribution Characteristics Based On Weibo Topics

Nowadays,with the rapid development of information technology,more and more data is accumulated in the Internet,and information flooding is becoming more and more serious.How to mine users' interesting topics from a large number of social media data has become an important issue to be solved in the information age.The emergence and development of probabilistic topic model represented by LDA provides an effective way to solve the mining and feature analysis of popular topics.As an important social media data source,micro-blog data has great application value in topic mining and feature analysis.In this paper,we use BTM model to extract topics and analyze the temporal and spatial characteristics of topics,the main research contents include:(1)The pre-processing of 2.92 million original microblog data yields about 900,000 valid data,then we get 48 topics using BTM model.The results are displayed in the form of word cloud and geographically visualized based on location attribute,and micro-blog topics are interlaced in space and the individual level is random.(2)In terms of spatial characteristics,by constructing the topic feature vector and using the cosine similarity to calculate the difference of the spatial distribution between the topics,it is found that the geographical distribution characteristics of the tourism topic are most obvious.According to the spatial distribution characteristics of the topics,six topics such as entrepreneurship,food,and transportation are selected as the research objects for spatial autocorrelation analysis,and the spatial scanning statistical method is used to detect hotspots for the topics with clustering mode;In term of temporal characteristics,by analyzing the variation trend of the number of microblogs over time,it is found that the number of topics related to entertainment and go sightseeing first increases and then decreases,while the number of topics related to trip and work first increases and then decreases.In terms of spatio-temporal characteristics,the similarity of provincial capital cities in daily life and vacation is measured by constructing urban feature vector and using cosine values.The result shows that the similarity between the provincial capitals is relatively high in daily life,but the difference between provincial capitals is highlighted during the holiday.By constructing interactive network of tourism topics,it is found that whether daily or holiday,Beijing and Shanghai are mainly outbound tourism,while Suzhou and Lijiang are mainly inbound tourism.(3)Pavement collapse is selected as a specific topic,and obtaining relevant microblog data through web spider.By using the BTM model to get the topic category of each weibo and introducing emotion attribute,the distribution characteristics of the collapse event in temporal and spatial are obtained.It is found that the attention degree of the road collapse event is decreasing year by year,and the subjective emotional intensity of people under various topics has a strong relationship with the number of incidents,but has a weak relationship with the number of microblogs,and the emotion is more likely to be incidents causing casualties.The analysis of weibo data is helpful to mine important topics and their evolution rules,which are of great significance for event detection,hot topic discovery,and public opinion analysis.