摘要
为提高城市恐怖威胁突发事件的监测预警效率,设计了一套基于改进型卷积神经网络(CNN)的恐怖威胁预警系统。系统对采集的恐怖威胁突发事件进行威胁度评估。相对于现有文本评估方法,提出结合改进型词频-逆文本频率的卷积神经网络用于评估威胁和监测预警。并通过对研判期间内事件的威胁度分析,划分了恐怖威胁等级。通过数据的可视化实现为相关部门提供监测预警信息。模型对比测试和实际运行结果表明,该系统相比于已有的CNN模型及区域型卷积神经网络(RCNN)模型,综合评估的精确度分别提升了5.4%和3%。
To improve the monitoring and early warning efficiency of urban terrorist threat emergencies,we designed a terror threat early warning system based on improved convolutional neural network(CNN).The system evaluates the threat degree of the collected terrorist threat emergencies.Compared with the existing text evaluation methods,a convolutional neural network based on the improved word frequency-inverse text frequency was proposed to evaluate threats and monitor early warning.Through the analysis of the threat degree of the events during the period of the study,the terror threat degree was divided,which can provide monitoring and early warning information for relevant departments through data visualization.Compared with the existing CNN model and region convolutional neural network model,the comprehensive evaluation accuracy of the system is improved by 5.4%and 3%,respectively.
作者
杜梦星
王彦伟
DU Mengxing;WANG Yanwei(School of Mechanical and Electrical Engineering,Wuhan Institute of Technology,Wuhan 430205,China)
出处
《武汉工程大学学报》
CAS
2020年第2期207-212,共6页
Journal of Wuhan Institute of Technology
基金
国家自然科学基金(51375186)
湖北省教育厅重点科研项目(D20161507)。