大数据下的销售异常发现与定位模型研究  

Research of Sales Anomaly Detecting and Locating Model Based on Big Data

作　　者：刘菊君 姜磊[1] 彭雄 周倩 杨先圣 LIU Ju-jun;JIANG Lei;PENG Xiong;ZHOU Qian;YANG Xian-sheng(School of Computer Science and Engineering, Hunan University of Science and Technology, Xiangtan 411201, China;BBK Commercial Chain Limited by Share Ltd, Xiangtan 411201, China)

机构地区：[1]湖南科技大学计算机科学与工程学院,湖南湘潭411201 [2]步步高商业连锁股份有限公司,湖南湘潭411201

出　　处：《小型微型计算机系统》2019年第1期64-68,共5页Journal of Chinese Computer Systems

基　　金：湖南省教育厅重点项目(15A064)资助;湖南省自然科学基金项目(2016JJ2056)资助

摘　　要：当今时代传统零售业竞争异常激烈且数据量庞大,因此在大数据平台下挖掘异常并让其辅助决策成为企业提高竞争力的有效手段.目前大多数离群点检测方法仅能对具有可比性的数据进行异常挖掘,但销售数据却受到季节性、节假日等因素影响而失去可比性,且管理层的需求并不仅仅是挖掘异常,其最终目的是定位异常、实现责任到人等实用意义,从而针对销售数据的异常发现与定位方法成为一大难题.为此提出了大数据下的销售数据的异常发现与定位模型.该模型利用权重的思想使数据具有可比性,从不同角度的数据进行离群点检测后通过建立概率模型实现异常定位.由于权重思想及独有的异常定位两个特征,该模型在实际应用于步步高商业连锁股份有限公司时获得了相关专业人员的高度认可.Nowadays,the competition of traditional retailing is very fierce and the amount of data is huge. So mining anomalies under the big data platform and then making auxiliary decision becomes an effective means for enterprises to improve their competitiveness.At present,most outlier detection can only mine the comparable data. However,sales data are affected by season,holiday and other factors and then lose comparability,and the requirement of management is not only mining anomalies,its ultimate purpose is to realize the practical significance such as locating anomalies and making it clear who is to blame. So the method of anomaly Detecting and Locating of sales data becomes an problem. Therefore,a model of Sales Anomaly Detecting and Locating Based on Big Data is proposed. The model makes the data comparable by using the idea of weight. The anomaly location is realized by establishing probability model after outlier detection from different angles. Due to characteristics of weight and unique anomaly location,the model is highly recognized by relevant professionals when it is applied to the BBK commercial chain Limited by Share Ltd.

关 键 词：大数据 离群点检测 异常发现 异常定位 权重 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]