适用于深度学习的公安情报结构化数据编码方法研究
|
摘要
[目的/意义]旨在为公安情报数据挖掘与定量分析提供参考。[过程/方法]阐述深度学习技术在非结构化数据中的成功应用,对比结构化数据与非结构化数据的差异,分析使用深度学习技术对公安情报结构化数据进行数据挖掘与定量分析时存在的主要障碍,提出基于实体嵌入的编码方法解决现有编码方法存在的向量稀松、难以计算等问题。[结果/结论]基于实体嵌入的编码方法可实现类别数据向数值数据的有效转换,进而解决公安情报结构化数据不适用于深度学习框架的问题。
[Purpose/significance]The paper is to provide references for mining and quantitative analysis of police intelligence data. [Method/process] The paper elaborates the successful application of deep learning technology in unstructured data,compares the differences between structured and unstructured data,analyzes the main obstacles when using deep learning techniques to conduct data mining and quantitative analysis of police information structured data,and proposes the Entity Embedding-based coding method to solve the problems of rarity of vectors and the difficulty of calculation.[Result/conclusion] The coding method based on entity embedding can realize the effective conversion of category data into numerical data,and then solve the problems that public security intelligence structured data is not suitable for deep learning framework.
[Purpose/significance]The paper is to provide references for mining and quantitative analysis of police intelligence data. [Method/process] The paper elaborates the successful application of deep learning technology in unstructured data,compares the differences between structured and unstructured data,analyzes the main obstacles when using deep learning techniques to conduct data mining and quantitative analysis of police information structured data,and proposes the Entity Embedding-based coding method to solve the problems of rarity of vectors and the difficulty of calculation.[Result/conclusion] The coding method based on entity embedding can realize the effective conversion of category data into numerical data,and then solve the problems that public security intelligence structured data is not suitable for deep learning framework.
引文
[1]彭知辉.论公安情报工作的性质、内容与作用[J].辽宁警专学报,2005,1(29):38.
[2]彭知辉.大数据:让情报主导警务成为现实[J].情报杂志,2015(5):1-6.
[3]傅子洋,徐荣贞,刘文强.基于贝叶斯网络的恐怖袭击预警模型研究[J].灾害学,2016(3):184-189.
[4]李勇男,蒋东龙,梅建明.基于基尼系数的决策树在涉恐情报分析中的应用[J].情报杂志,2017,(4):29-33.
[5]张强,何乐平.基于支持向量机的治安高危人员风险预警办法[J].管理评论,2016(11):245-251.
[6]马素华.随机森林模型在评估个人危险性方面的应用[J].现在侦查技战法论坛,2017(1):219-226.
[7]梅毅,熊婷,蒋德荣.网络中涉恐信息的高效挖掘模型研究与仿真[J].计算机仿真,2015(5):302-305,314.
[8]曲晓峰.人工智能、机器学习和深度学习之间的区别和联系[EB/OL].[2016-12-20].http://www.leiphone.com/news/201609/gox8CoyqMrXMi4L4.html.
[9]山世光,阚美娜,刘听,等.深度学习:多层神经网络的复兴与变革[J].科技导报,2016,34(14):60-70.
[10]卢宏涛,张秦川.深度卷积神经网络在计算机视觉中的应用研究综述[J].数据采集与处理,2016(1):2-15.
[11]CHEN J,KANG X,LIU Y,et al.Median filtering forensics based on convolutional neural networks[J].Signal Processing Letters IEEE,2015,22(11):1849-1853.
[12]LOPES N,RIBEIRO B.Towards adaptive learning with improved convergence of deep belief networks on graphics processing units[J].Pattern Recognition,2014,47(1):114-127.
[2]彭知辉.大数据:让情报主导警务成为现实[J].情报杂志,2015(5):1-6.
[3]傅子洋,徐荣贞,刘文强.基于贝叶斯网络的恐怖袭击预警模型研究[J].灾害学,2016(3):184-189.
[4]李勇男,蒋东龙,梅建明.基于基尼系数的决策树在涉恐情报分析中的应用[J].情报杂志,2017,(4):29-33.
[5]张强,何乐平.基于支持向量机的治安高危人员风险预警办法[J].管理评论,2016(11):245-251.
[6]马素华.随机森林模型在评估个人危险性方面的应用[J].现在侦查技战法论坛,2017(1):219-226.
[7]梅毅,熊婷,蒋德荣.网络中涉恐信息的高效挖掘模型研究与仿真[J].计算机仿真,2015(5):302-305,314.
[8]曲晓峰.人工智能、机器学习和深度学习之间的区别和联系[EB/OL].[2016-12-20].http://www.leiphone.com/news/201609/gox8CoyqMrXMi4L4.html.
[9]山世光,阚美娜,刘听,等.深度学习:多层神经网络的复兴与变革[J].科技导报,2016,34(14):60-70.
[10]卢宏涛,张秦川.深度卷积神经网络在计算机视觉中的应用研究综述[J].数据采集与处理,2016(1):2-15.
[11]CHEN J,KANG X,LIU Y,et al.Median filtering forensics based on convolutional neural networks[J].Signal Processing Letters IEEE,2015,22(11):1849-1853.
[12]LOPES N,RIBEIRO B.Towards adaptive learning with improved convergence of deep belief networks on graphics processing units[J].Pattern Recognition,2014,47(1):114-127.