摘要
准确估算洪涝应急物资需求是城市公共安全应急管理的重要组成部分。提出了一种基于百度大数据统计人口动态数量的方法,将人口动态数量、季节系数和地区系数作为输入变量,运用极限学习机方法构建了洪涝灾害应急物资需求估算模型。以2019年5月25日武汉市发生特大暴雨事件为例,分别估算当天7时、14时和21时的中心城区应急物资需求量。结果表明,利用百度大数据能够统计人口动态数量,通过极限学习机所构建的洪涝灾害应急物资估算模型能够动态估算实时应急物资需求量。
The present paper is aimed at proposing a so-called extreme learning machine method to solve the problem for estimation of the flood emergency material demand instead of the traditional census and the traditional spatial distribution statistic methods. For,the above two methods fail to reflect the dynamic population distribution in the urban public safety emergency management in case of the non-linear and irregular demand for the flood saving material. The reason of our choice lies in that the estimation of the emergency material demand is closely related to the actual situation forecast and the number of the people actually affected. Such a choice has also been done is due to the readiness to obtain the heat map on the dynamic population. As a kind of entirely new type of neural network,the map enjoys extinguished performance in solving the complicated non-linear data fitting by using the dynamic changes of the massive data amount to gain the global optimized solutions. For instance,the system can take 3 indices as the input variables,i. e. the dynamic population,the seasonal coefficient and the regional coefficient.When it is necessary to build up an estimation cluster of data and information of the urban flood emergency material demand( e. g.the needed quantity of tents,quilts,and folding beds),the extreme learning machine method can also be adopted by way of drawing out similar historical statistical records for the flood disaster events both at home or abroad as a training sampling set. In such a case,3 available indices can be taken to test the simulation accuracy of the actual and/or forecast data concerned,including the root mean square error,the model validity coefficient and the decisive factors in the simulated modeling process. And,then,such 3 available indices can be put into hypothetical assessment so that the model can be made up to display its utmost reliability. And,so,let’s take the torrential rainstorm in Wuhan that took place on May 25,2019 as an example,the real-time population involved at
作者
吴浩
常春慧
昝军
刘毅
李珍
WU Hao;CHANG Chun-hui;ZAN Jun;LIU Yi;LI Zhen(School of Safety Science and Emergency Management,Wuhan University of Technology,Wuhan 430070,China;College of Urban and Environmental Sciences,Central China Normal University,Wuhan 430079,China;Hubei Provincial Safety Production Emergency Rescue Center,Wuhan 430070,China)
出处
《安全与环境学报》
CAS
CSCD
北大核心
2020年第5期1829-1835,共7页
Journal of Safety and Environment
基金
国家重点研发计划项目(2018YFC0810600)
国家自然科学基金面上项目(41671406)。
关键词
公共安全
应急物资
大数据应用
极限学习机
public safety
emergency materials
big data applications
extreme learning machine
