基于微博数据的台风“山竹”舆情演化时空分析

Spatial and temporal analysis of network public opinion evolution of typhoon“Mangkhut”based on Weibo data

将情感分析模型、动态演化模型、话题聚类模型、网络社团模型结合地理可视化技术应用到台风的灾害评估中。将微博情绪与台风灾害联系起来,从情感值与讨论热度两个角度入手,根据台风"山竹"相关话题的25798条微博数据,完整的展示本次事件网络舆情的演化过程,通过隐含狄利克雷分布(latent dirichlet allocation,LDA)主题模型挖掘用户关注话题,发现台风登陆事件与湖南收费站事件对公众情绪的消极影响;抽取台风"山竹"相关微博中蕴含的地理位置信息,建立广东省21个城市的网络社团模型,检验用户情绪、城市词频、用户位置、网络节点活跃度等指标探测受灾城市的能力;根据广东省38个气象站点的24 h最大降雨数据进行空间插值。降水主要集中在广东南部地区,阳江市发生特大暴雨,引发了严重的洪涝灾害,其情绪值也是最低的。

Internet public opinion was the sum of the public various emotions,attitudes and opinions on related topics.This paper applied the sentiment analysis model,dynamic evolution model,topic clustering model,network community model and geographic visualization technology to the typhoon disaster assessment.This research fully illustrated how the public opinion on typhoon"Mangkhut"evolved by analyzing the 25798 Weibo related from the two perspectives of emotional value and discussion heat.By utilizing LDA clustering method the negative impacts of typhoon"Mangkhut"′s landing and"Hunan toll station event"on public sentiment were found.After collecting the geographical location information of those typhoon"Mangkhut"related Weibo,a network-community model of 21 cities in Guangdong province was established and the model tested the ability to explore the affected cities through such indicators as users′sentiments,city word frequency,users′location,and network node activity.Spatial interpolation was performed based on the 24 h maximum rainfall data from 38 meteorological stations in Guangdong province.Precipitation was mainly concentrated in the southern part of Guangdong.Heavy rains occurred in Yangjiang City,which caused severe flooding and the lowest emotional value.