机动车辆保险欺诈检测系统及团伙识别研究

发布时间：2018-07-08 13:41

  本文选题：保险欺诈 + 矩阵排序　； 参考：《保险研究》2017年02期

【摘要】：随着我国保险业的蓬勃发展,车险领域的保险欺诈问题日益严峻。鉴于我国在机动车辆保险欺诈检测技术方面较为滞后,本文针对车险反欺诈检测方法进行研究,首次将团伙微观建模应用于机动车辆保险欺诈检测。通过引入广义团伙概念,采用基于矩阵的相似度计算、秩排序和变换算法,对极小概率发生但又高度可疑的团伙实现有效识别。相较于传统方法具有更准确和高效的实际应用价值:引入广义团伙对车险欺诈进行全方位识别;将可疑欺诈团伙的车辆碰撞关系映射为人网络关系,从而避免各种人为规避行为对识别和检测的影响;不需要确定的欺诈样本,也不需要进行模型训练就可以直接应用;采用矩阵数值运算完成全部过程,有效提高计算效率。
[Abstract]:With the rapid development of insurance industry in China, insurance fraud in auto insurance field is becoming more and more serious. In view of the lag in the detection technology of motor vehicle insurance fraud in China, the anti-fraud detection method of vehicle insurance is studied in this paper, and the group microscopic modeling is applied to the detection of fraud in motor vehicle insurance for the first time. By introducing the concept of generalized group, using the similarity calculation based on matrix, rank ranking and transformation algorithm, we can effectively identify the group with minimal probability but highly suspicious. Compared with the traditional method, it has more accurate and efficient practical application value: introducing the generalized group to identify the vehicle insurance fraud in all directions, mapping the vehicle collision relation of the suspected fraud group to the network relationship, In order to avoid the impact of various human evading behavior on identification and detection; do not need to determine the sample fraud, and do not need model training to be directly applied; the use of matrix numerical operation to complete the whole process, effectively improve the efficiency of the calculation.
【作者单位】： 西安交通大学经济与金融学院;深圳般若计算机系统股份有限公司;太平财产保险有限公司;
【分类号】：F842.634;D924.35
，

本文编号：2107619