CSV框架下欺诈索赔的筛选与审查策略研究

发布时间：2018-11-11 11:46

【摘要】：本论文依据保险欺诈研究中的CSV(costly state verification)理论框架,从保险公司欺诈索赔管理的现实背景出发,构造了欺诈索赔检查的双阶段流程,在此基础上建立了相应的最优检查模型及特征指标识别模型。最优检查模型的结论表明,现实中十分普遍的选择性检查并不总是最好的选择;某些情况下,全面检查方案或不检查方案可成为更合适的代替方案。在最小化公司总损失的目标下,全面检查方案具有关于审查成本的最优检查策略,而选择性检查方案则具有关于审查比例和审查成本的最优检查策略组合。论文利用MATLAB仿真模拟,分析了模型参数如欺诈率、欺诈索赔额、筛选效果、审查效果等对各检查方案及其最优策略的影响。审查比例是双阶段检查模型中的重要变量,它决定着检查方案的可行性。成功筛选出欺诈嫌疑程度较高的索赔,并将之送入审查阶段,既有助于节省检查成本,也有助于提高检查效率。关于筛选过程,本论文建立了可疑索赔识别的特征指标模型,并利用某保险公司的车险索赔数据对识别模型进行了适用性验证。所用数据全部来源于保险公司的真实索赔案件统计库,特征指标则源于保险公司内部系统针对欺诈嫌疑设置的经验性识别因子。结果表明,构建合适的特征指标体系可以有效筛选出欺诈嫌疑较大的索赔,而筛选模型的预测效果则依赖于保险公司索赔数据记录的完整性。论文主要贡献包括三个方面。一是将保险欺诈研究的经典CSV理论框架与保险公司索赔管理的现实流程相结合,建立了欺诈索赔检查的双阶段流程模型。二是利用筛选模型和检查模型分别描述该流程中嫌疑欺诈索赔的筛选、审查两个阶段,以获得一定条件下的最优检查策略。三是研究结论可为保险公司制定欺诈索赔管理策略、实现总成本最小化目标提供必要的理论依据与方法支持。
[Abstract]:According to the CSV (costly state verification) theory framework in the study of insurance fraud and the realistic background of fraud claim management in insurance companies, this paper constructs a two-stage process of fraud claim inspection. On this basis, the corresponding optimal checking model and characteristic index recognition model are established. The conclusion of the optimal checking model shows that the universal selective checking is not always the best choice, and in some cases, the overall checking scheme or the non-checking scheme can be a more suitable alternative. Under the objective of minimizing the total loss of the company, the overall inspection scheme has an optimal inspection strategy on the audit cost, while the selective inspection scheme has an optimal combination of the inspection strategy on the audit ratio and the audit cost. In this paper, MATLAB simulation is used to analyze the influence of model parameters such as fraud rate, amount of fraud claim, screening effect and checking effect on each checking scheme and its optimal strategy. The audit scale is an important variable in the two-stage inspection model, which determines the feasibility of the inspection scheme. The successful screening of claims with high level of fraud suspicion and sending them to the examination stage will not only save the cost of inspection but also improve the efficiency of inspection. Regarding the screening process, this paper establishes the characteristic index model of suspicious claim identification, and verifies the applicability of the model by using the vehicle insurance claim data of a certain insurance company. All the data used are from the real claim statistics database of insurance companies, and the characteristic indicators are derived from the empirical identification factors set up by the internal system of insurance companies for suspected fraud. The results show that a suitable characteristic index system can be used to screen claims with high fraud suspicion, and the prediction effect of the screening model depends on the integrity of claim data records of insurance companies. The main contributions of the thesis include three aspects. The first is to combine the classical CSV theory framework of insurance fraud research with the real process of claim management in insurance companies, and establish a two-stage process model for fraud claim inspection. The second is to use the screening model and the check model to describe the selection of the suspected fraud claims in the process, and examine the two stages to obtain the optimal check strategy under certain conditions. Third, the conclusion can provide the necessary theoretical basis and method support for insurance companies to formulate fraud claim management strategy and achieve the goal of total cost minimization.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：D924.35;F842

【相似文献】