基于旋转森林集成学习的涉恐实体挖掘研究  

Research of an Rotation Forest based on Ensemble Classifier in Terrorists Entity Mining

作　　者：孙菲菲[1] 林平[1] 曹卓[2] 

机构地区：[1]中国人民公安大学反恐怖学院,北京100038 [2]中国人民公安大学犯罪学院,北京100038

出　　处：《情报杂志》2015年第5期190-195,共6页Journal of Intelligence

摘　　要：近期,我国出现了一系列暴力恐怖袭击事件,面临的反恐形势日趋严峻。传统的被动警务防范模型弊端日益凸显,亟需有效应对恐怖袭击的智能化防范策略。作为近年来的研究热点,数据挖掘技术和应用发展迅速。文章充分利用数据挖掘的前沿技术和成果,提出了一个相对完整的涉恐实体挖掘模型,以期为反恐实战工作服务。其中,简要介绍针对中文和维吾尔文非结构化文本数据的挖掘流程,提出一个有效应用于识别和预测涉恐实体的基于旋转森林的集成分类模型,并用实验证明这一模型较之一般的分类器会有更优越的分类性能。Recently, China experienced a series of violent terrorism attacks incidents. Faced with an increasingly grim situation of terror, China is lagging behind counter-terrorism measures. It leads to taking intelligence measures to deal with terrorism attack effectively. Data miningbecomes a hot issue and grows rapidly in recent years. In this paper, we present a relatively complete model to direct counter-ter- rorism war with the help of the advanced technology and achievement of data mining. Firstly, it describes briefly the process of the text mining procedure using Chinese and Uygur language. Secondly, we introduce a new way of classification, which can be applied to terrorist identification. Finally it demonstrates the high reliablity regarding classifier results compared to other classifiers.

关 键 词：反恐 非结构化文本 数据挖掘 旋转森林 集成学习 

分 类 号：G353[文化科学—情报学]