基于深度学习和公开来源信息的反恐情报挖掘

发布时间：2019-03-30 21:58

【摘要】：[目的/意义]反恐情报工作通常依赖于情报人员的智力和经验,但大数据时代对海量数据的处理仅仅依靠人力显然是不够的。为了提高反恐情报信息工作效率,从而全面提升反恐情报预警能力,迫切需要自动化智能化的情报处理技术来减轻情报人员的工作量。[方法/过程]利用人工智能领域最新成果深度学习技术对公开来源信息进行反恐情报挖掘和分类,并采用深度学习词向量工具Word2vec对大量开源中文语料库进行深度训练。[结果/结论]实验结果证明深度学习算法在反恐情报的挖掘分类精度上要优于传统情报分类算法;通过Word2vec计算相似词向量的余弦距离,利用已知的恐怖分子、恐怖组织名单来挖掘与其密切关联的潜在恐怖分子,对于反恐情报预警工作具有积极意义。[局限]深度学习算法需要大量样本进行训练,时间开销稍大。
[Abstract]:[purpose] counter-terrorism intelligence usually depends on the intelligence and experience of intelligence personnel, but the processing of massive data in big data's era is obviously not enough by manpower alone. In order to improve the efficiency of counter-terrorism intelligence information and improve the ability of anti-terrorism intelligence early warning, automatic intelligence processing technology is urgently needed to reduce the workload of intelligence personnel. [methods / process] Anti-terrorism intelligence mining and classification of open-source information are carried out by using the latest in-depth learning technology in the field of artificial intelligence, and in-depth learning word vector tool Word2vec is used to train a large number of open-source Chinese corpus in depth. [results / conclusion] the experimental results show that the depth learning algorithm is superior to the traditional information classification algorithm in the classification accuracy of counter-terrorism intelligence mining. The calculation of cosine distance of similar word vectors by Word2vec and the use of known lists of terrorists and terrorist organizations to excavate potential terrorists closely associated with them are of positive significance for the work of anti-terrorism intelligence early warning. [limitation] the deep learning algorithm needs a large number of samples to be trained, and the time cost is a little large.
【作者单位】：武警警官学院;国防科技大学信息通信学院;
【基金】：国家社会科学基金西部项目“基于信息网络的反恐预警体系架构及关键问题研究”(项目编号:16XSH003);国家社会科学基金军事学项目“武警部队基于信息网络的反恐预警研究”(项目编号:16GJ003-245)的阶段性成果
【分类号】：G350

【相似文献】