地球信息科学学报 ›› 2018, Vol. 20 ›› Issue (6): 807-816.doi: 10.12082/dqxxkx.2018.180022.

• 2017年中国地理信息科学理论与方法学术年会优秀论文专辑 • 上一篇    下一篇

社交媒体数据对反映台风灾害时空分布的有效性研究

梁春阳1(), 林广发1,2,3,*(), 张明锋1,2,3, 汪玮杨1, 张文富1, 林金煌1, 邓超1   

  1. 1. 福建师范大学 地理研究所,福州 350007
    2. 福建省陆地灾害监测评估工程技术研究中心,福州 350007
    3. 海西地理国情动态监测与应急保障研究中心,福州 350007
  • 收稿日期:2018-01-02 修回日期:2018-04-08 出版日期:2018-06-20 发布日期:2018-06-20
  • 通讯作者: 林广发 E-mail:pepsi8696@163.com;GuangfaLin@qq.com
  • 作者简介:

    作者简介:梁春阳(1993-),男,硕士生,研究方向为自发地理信息与应急管理。E-mail: pepsi8696@163.com

  • 基金资助:
    国家重点研发计划重点专项(2016YFC0502905);福建省公益科研院所专项(2015R1034-1);福建省测绘地理信息局科技资助项目(2017JX03)

Assessing the Effectiveness of Social Media Data in Mapping the Distribution of Typhoon Disasters

LIANG Chunyang1(), LIN Guangfa1,2,3,*(), ZHANG Mingfeng1,2,3, WANG Weiyang1, ZHANG Wenfu1, LIN Jinhuang1, DENG Chao1   

  1. 1. Institute of Geography, Fujian Normal University, Fuzhou 350007, China
    2. Fujian Provincial Engineering Research Center for Monitoring and Assessing Terrestrial Disaster, Fuzhou 350007, China
    3. Research Center for National Geographical Condition Monitoring and Emergency Support in the Economic Zone on the West Side of the Taiwan Strait, Fuzhou 350007, China
  • Received:2018-01-02 Revised:2018-04-08 Online:2018-06-20 Published:2018-06-20
  • Contact: LIN Guangfa E-mail:pepsi8696@163.com;GuangfaLin@qq.com
  • Supported by:
    National Key Research and Development Program of China, No.2016YFC0505905;Non-profit Research Projects of Fujian Province, No.2015R1034-1;Development Foundation of Surveying, Mapping and Geoinformatics of Fujian Province, No.2017JX03

摘要:

当灾害事件发生时,与之相关的社交媒体数据不断产生,其中包含了丰富的灾情信息和签到地理位置信息,这为灾情态势的及时感知提供了一种新的数据源,但是因社交媒体用户量的地区差异及网络空间中信息传播模式的特点,给社交媒体签到数据所代表的空间点过程的模式分析带来了一些新的问题,如签到点密度与实际灾害点事件密度之间的对应关系、签到点之间的空间关系、点格局的空间异质性及其影响因素等。本文以2016年14号台风“莫兰蒂”为例,以“台风”和“莫兰蒂”为关键词,在新浪微博平台上采集了2016年9月14-17日的微博数据,使用文档主题生成模型(Latent Dirichlet Allocation, LDA)和支持向量机(Support Vector Machine, SVM)对微博文本进行分类,构建了含有签到位置信息的灾情点事件数据库。在此基础上,针对社交媒体用户分布的空间异质性提出了一种基于签到点用户活跃度的加权模型。以全局自相关统计量Moran′s I为指标,对加权前后的签到微博数据进行对比,发现这些在社交网络中产生的签到微博数据在现实地理空间中存在明显的空间自相关性;基于“雨”、“停电”等关键词,利用上述加权处理后的微博数据库进行灾害制图,结合真实灾情资料进行时空对比分析,结果表明系列图谱能够反映台风灾害的时空过程趋势。

关键词: 社交媒体, 台风灾害, 空间分析, 数据挖掘, 空间自相关

Abstract:

When a disaster occurs, a large number of images and texts with geographic information quickly flood the social network, which provides a new data source for timely awareness of disaster situations. However, due to the regional variation in the number of social media users and characteristics of information diffusion in cyberspace, new problems have risen in the mode analysis of spatial point processes represented by the check-in data. Examples are the correlation between check-in point density and disaster location density, spatial relation between check-in points or spatial heterogeneity of point pattern and associated influences. In this study, we took Typhoon No.14 in 2016 as an example and collected Sina Weibo data between September 14 and September 17, 2016 using keywords “Typhoon” and “Meranti”. We classified the Weibo texts using Latent Dirichlet Allocation (LDA) and Support Vector Machine (SVM) algorithms and constructed a disaster database containing relevant check-in information. In addition, considering the spatial heterogeneity of Weibo users, we proposed a weighted model based on user activity at the check-in points. Using the global autocorrelation statistics Moran′s I as an indicator, we compared the check-in data before and after adding weights and discovered obvious spatial autocorrelation of the check-in data in real geographical locations. We tested our model on Weibo data with keyword “rain” and “power failure”. The results show that a series of maps generated by our model is able to reflect the typhoon disaster spatio-temporal process trends.

Key words: social media, typhoon disaster, spatial analysis, data mining, spatial autocorrelation