| 基于深度学习的多维情景空间下群体性事件分析与预测研究 |
| |
| 引用本文： | 张鼎华,李卫俊,李丞,申世飞. 基于深度学习的多维情景空间下群体性事件分析与预测研究[J]. 中国管理科学, 2020, 28(8): 172-180. DOI: 10.16381/j.cnki.issn1003-207x.2020.08.015 |
| |
| 作者姓名： | 张鼎华 李卫俊 李丞 申世飞 |
| |
| 作者单位： | 1. 华南理工大学公共管理学院, 广东 广州 510641;2. 华南理工大学地方风险治理研究中心, 广东 广州 510641;3. 清华大学公共安全研究院, 北京 100084 |
| |
| 基金项目： | 中央高校基本业务费资助项目（2018ZDXM15）；广州市哲学社科规划项目（2018GZYB18）；广东省哲学社科规划项目（GD17CGL04）；教育部重大攻关项目（16JZD026）；国家自然科学基金重大研究计划资助项目（91024016） |
| |
| 摘 要： | 在社会转型的背景下,我国群体性事件的数量不断攀升,极大地影响了社会的和谐稳定。本文应用"情景-次级情景-对象-要素"模型对各类群体性事件案例进行分解,提取相关的影响因素,构建多维情景空间模型,并提出了用于群体性事件预测的卷积神经网络模型,阐述了其基本原理,然后结合实例说明了其应用。通过根据多维情景空间模型对群体性事件案例进行编码形成的数据样本集,来训练该预测模型,并用AUC值评估其有效性。最后分析了不同影响因素在群体性事件预测中的作用以及应急管理主体的应对方向。
|
| 关 键 词： | 群体性事件 多维情景空间模型 卷积神经网络 预测 应急管理 |
| 收稿时间： | 2017-12-20 |
| 修稿时间： | 2018-04-08 |
The Researchon the Analysis and Prediction of Mass Incidentsin Multi-dimensional Scenario Space Based on Deep Learning |
| |
| ZHANG Ding-hua,LI Wei-jun,LI Cheng,SHEN Shi-fei. The Researchon the Analysis and Prediction of Mass Incidentsin Multi-dimensional Scenario Space Based on Deep Learning[J]. Chinese Journal of Management Science, 2020, 28(8): 172-180. DOI: 10.16381/j.cnki.issn1003-207x.2020.08.015 |
| |
| Authors: | ZHANG Ding-hua LI Wei-jun LI Cheng SHEN Shi-fei |
| |
| Affiliation: | 1. School of Public Administration, South China University of Technology, Guangzhou 510641, China;2. Research Center of Local Risk Management, South China University of Technology, Guangzhou 510641, China;3. Institute of Public Safety, Tsinghua University, Beijing 100084, China |
| |
| Abstract: | The increasing mass incidents have greatly affected the harmony and stability of the society in the context of social transformation. A multi-dimensional scenario space model is constructed by using the "Scenario-Sub-Scenario-Object-Factor" model to decompose of all kinds of mass incidents and extract related influential factors. Based on multi-dimensional scenario space model, the convolutional neural network model is applied to mass incidents prediction, its principle is explained in detail and practical applications are discussed. A data set, formed by encoding a group of mass incidents' cases based on the multidimensional scenario space model is used to train (test) the predictive model and evaluat its validity via Area Under Curve (AUC). Furthermore, the effect of different factors on the prediction of mass incidents is analyzed and the direction of emergency management is indicated. |
| |
| Keywords: | mass incidents multi-dimensional scenario space model convolutional neural network prediction emergency management |
| 本文献已被 维普 等数据库收录！ |
| 点击此处可从《中国管理科学》浏览原始摘要信息 |
|
点击此处可从《中国管理科学》下载全文 |
|
