孤立点挖掘在警情时间序列异常点识别中的应用  

The Application of Outlier Mining Algorithm in Exceptional Crime Time Series Identification

作　　者：陈鹏[1] 胡啸峰[1] 林艳[1] 

机构地区：[1]中国人民公安大学警务信息工程学院,北京102600

出　　处：《科学技术与工程》2015年第7期225-228,233,共5页Science Technology and Engineering

基　　金：国家"十二五"科技支撑计划项目(2013BAK02B05);中国人民公安大学基科费项目(2014JKF01118)资助

摘　　要：针对北京市的警情时间序列中存在的一些异常特征,采用了一种基于距离和的孤立点挖掘方法来进行分析。该方法将警情在每日不同时段内的分布处理为相应的属性,然后将警情的时段属性进行标准化,计算了不同点之间的欧式距离和来判断点的异常性。结果发现,北京的警情异常现象出现在春节和年底期间,其中前者表现为由于居民燃放烟花爆竹所引起的火灾事故和治安纠纷;而后者则表现为电信诈骗案的高发。与基于日接警数量的异常点分析相比,基于距离和的异常点挖掘能够有效地将警情在时序分布中隐藏的一些异常聚集效应识别出来,从而为公安部门开展预警和情报信息服务提供帮助。An outlier-based mining method is used to analyze the exceptional points in time series of crime in Beijing in China. Firstly, the crime counts in different hours of day are treated as the attributes, and then the exceptional character is estimated by computing the Euclidean distance between each pair of data whose temporal attributes are standardized. The findings indicate that the exceptional clustering of crime concentrate in Spring Festival and year＇ s end. The phenomenons are due to fire disaster and disputing cause by fireworks and telecommunications fraud. Comparing to the old way of data processing that identifies the exceptional point by counting crime frequency, the distance-based outlier mining method could find the unusual pattern of crime clustering in more effec- tive way. Therefore, this method could be used in helping police providing the service of early warning and informa- tion analysis.

关 键 词：孤立点挖掘 时间序列 警情预测 

分 类 号：TP391.75[自动化与计算机技术—计算机应用技术]