一种基于快递大数据的异常检测系统

An anomaly detection system based on express big data

随着信息时代的到来,快递行业迅速发展起来,推动着流通方式的转型和消费升级。人们在享受快递业发展带来巨大便捷的同时,也伴随着难以控制的流动性风险,给公共安全带来严峻的挑战。例如,偷窃的赃物通过快递方式进行销赃,利用快递方式运输毒品、爆炸物等危险物品。基于以上考虑,通过分析真实的历史快递记录,着力于研究利用快递方式进行销赃这一类犯罪行为,进而以识别该类嫌疑人为研究目标,从统计、时间和地理3方面特征进行了详细的分析。另外,提出了一种Two-Step异常检测方法用于嫌疑人的识别。该方法分为2步,第1步是过滤正常用户,第2步是识别嫌疑人。实验结果表明,通过该方法能够准确地识别出嫌疑对象,相比较传统方法,该方法能够有效地解决正负类数据不平衡问题,降低误检率,因此具有较高的实用价值。

With the advent of the information age,the express delivery industry has developed rapidly,which has promoted the transformation of circulation methods and the consumption upgrading.While enjoying the tremendous convenience of the development of the express delivery industry,people are also facing uncontrollable liquidity risks,and serious challenges are posed to public safety.For example,stolen goods are sold by express delivery,and dangerous goods such as drugs and explosives are transported by express delivery.Given the abovementioned considerations,we focus on the crimes of using express delivery to deal with stolen goods based on the analysis of actual historical records of express delivery,and then take the identification of such criminal suspects as the research target to conduct detailed analysis from the aspects of statistics,time,and geography.In addition,we also present a two-step anomaly detection method for suspect identification.The first step is to filter normal users,and the second step is to identify suspects.Experimental results show that in comparison with traditional methods,this method can accurately identify suspects,effectively solve the problem of positive and negative data imbalance,and significantly reduce false detection rate.Therefore,it has high practical value.