ATM交易成功率的异常检测报警与选址  

Abnormity Alarming for Transaction Success Rate of ATM and Site Selection

作　　者：冯涛 曲宇勋 丁山 王铮涛 袁博 张洪彬 FENG Tao;QU Yuxun;DING Shan;WANG Zhengtao;YUAN Bo;ZHANG Hongbin(Qiushi Honors College,Tianjin University,Tianjin 300072,China;College of Management and Economics,Tianjin University,Tianjin 300072,China)

机构地区：[1]天津大学求是学部,天津300072 [2]天津大学管理与经济学部,天津300072

出　　处：《数学建模及其应用》2018年第3期68-78,F0002,F0003,共13页Mathematical Modeling and Its Applications

摘　　要：ATM交易故障的监测与解决是商业银行运营中的难题,研究ATM数据中交易成功率的检异报警机制与ATM选址,对于解决上述问题具有重要意义.基于上述考虑,首先建立基于中心极限定理的方差标准化数据校正模型,消除交易量对成功率的平均效应;其次,将处理后的成功率数据导入孤立森林算法,输出数据的异常度;之后,引入K-means算法对异常程度进行分级,提取异常点;最后,引入衰变连续函数报警系统,得到报警结果.本文又从资源利用率最大化的角度出发,结合移动电子支付对于ATM需求的影响,建立ATM全局分布最优模型,选择合理的ATM分布方案;并将选址结果与运营状况良好的ATM实际地址进行比较,验证模型的合理性及优越性.On the basis of previous researches,the paper made progress and proposed solution mainly on the alarm system about the occurrence of the abnormal transaction success rate of ATM. Variance-standardized data correction model was firstly adopted. Besides, Isolation Forest, K-means, as well as decayed continuous function were also effectively employed in the article to solve corresponding problems such as average effect. The paper used the degree of coverage and activity to select sites for ATM. Besides, comparison with actual locations of ATM further proved the validity of the site selection model.

关 键 词：平均效应 孤立森林 K-MEANS 覆盖度 活跃度 

分 类 号：O29[理学—应用数学]