P公司互联网支付实时风险控制的架构方案设计

发布时间：2018-03-22 19:27

本文选题：实时风控　切入点：数据挖掘　出处：《东华大学》2017年硕士论文　论文类型：学位论文

【摘要】：互联网支付现已成为我们生活中必不可少的组成部分,然而互联网支付并非是安全的。由于风险事件频发,支付风险与日俱增,严重影响了用户的支付体验。P公司作为一家规模较大的第三方支付公司,在风险控制领域积累了一定的经验。P公司现有的风控体系主要是采用事后风控的方式。在发生风险事件后,主要在资金结算或资金出款的时候作拦截,能够阻止大部分风险事件的发生。但是,该风控体系不是全流程风控,用户体验差,实时风控力度不够,容易让犯罪分子有可乘之机。本课题基于实时风控理论,利用了数据挖据算法和云计算技术,进行了P公司实时风险控制的架构方案设计并应用于P公司现有的风控体系中。首先,本课题从实时风险管理的理论出发,论证了现有风控体系的弊端,结合P公司的风险场景,提出了以实时风控为主、准实时风控和事后风控为辅的全流程风控体系。其中,实时风控的风控类型包括商户风控、订单风控、用户风控和交易风控;通过把实时风控中耗时的业务放到准实时风控中,加快了实时风控的响应速度;通过在事后风控中应用数据挖掘算法,把商户、用户和交易的风险等级分类,预测商户、用户和交易未来的风险等级,并将预测的风险等级应用于实时风控的规则控制模块。其次,在应用架构上,本课题把实时风控系统划分成8个子系统,分别是实时风控SOA子系统、规则引擎子系统、准实时风控子系统、数据采集子系统、数据仓库子系统、风险评级子系统、风控监控子系统和风控运营子系统。各子系统间互相协作、互相配合,共同组成了实时风控系统。在技术层面,本课题研究了云计算技术,全新设计了数据访问框架、SOA框架、消息队列框架和自动扩容系统。在网络架构上,本课题使用了两地四中心的多IDC解决方案,应用部署采用了虚拟化技术。在数据库架构上,进行数据库的水平和垂直拆分。在性能架构上,提出了单场景压力测试和全场景稳定性测试,实时风控SOA接口的生产指标为720TPS/0.2s,须满足7*24小时的稳定性测试要求。通过这些技术措施,增加了实时风控的实时性和可靠性,为实时风控起到了强有力的技术支撑。本课题亦研究了实时风控的战略管理和运营管理。通过实时风控的SWOT分析,分析了P公司实时风控体系的优势和劣势、机会和威胁。运用互联网运营的4P方法,研究了P公司的实时风控运营体系的建立以及运营团队的建设。最后,通过P公司实时风控的实施效果分析,阐述实时风控的实际风险结果,分析该新系统尚存在的不足和改进建议。
[Abstract]:Internet payment has become an essential part of our lives, but Internet payment is not safe. Seriously affected the user's payment experience. P as a larger third-party payment company, Has accumulated some experience in the field of risk control. The existing risk control system of P Company mainly adopts the method of risk control afterwards. After a risk event, it mainly intercepts when the funds are settled or released. It can prevent the occurrence of most risk events. However, the risk control system is not a full-process risk control system, the user experience is poor, the real-time risk control is not strong enough, it is easy for criminals to take advantage of the opportunity. This topic is based on the theory of real-time risk control. Based on the data mining algorithm and cloud computing technology, the paper designs the real-time risk control architecture of P Company and applies it to the existing risk control system of P Company. Firstly, this paper starts from the theory of real-time risk management. The disadvantages of the existing risk control system are demonstrated. Combined with the risk scenario of P Company, a full-process risk control system with real-time risk control, quasi-real-time risk control and post-event risk control is put forward. Among them, the risk control types of real-time risk control include merchant risk control. Order risk control, user risk control and transaction risk control; by putting the time-consuming business in real-time risk control into quasi-real-time risk control, the response speed of real-time risk control is accelerated; and by applying data mining algorithm in post-event risk control, the merchants, Classification of risk levels of users and transactions, prediction of future risk levels of merchants, users and transactions, and application of the predicted risk levels to the rule control module of real-time risk control. Secondly, in the application architecture, This paper divides the real-time risk control system into eight subsystems, which are real-time risk control SOA subsystem, rule engine subsystem, quasi-real-time risk control subsystem, data acquisition subsystem, data warehouse subsystem, risk rating subsystem. The risk control and monitoring subsystem and the wind control operation subsystem. The subsystems cooperate with each other and form a real-time risk control system. At the technical level, this paper studies cloud computing technology, and designs a new data access framework SOA framework. Message queue framework and automatic expansion system. In the network architecture, this topic uses the multi-#en0# solution of four centers in both places, and the virtualization technology is used in the application deployment. In terms of performance architecture, single scenario stress test and full scenario stability test are proposed. The production index of real-time risk control SOA interface is 720TPS / 0.2s, which must meet the requirements of 72TPS / 0.2s stability test. Through these technical measures, the real-time and reliability of real-time risk control are increased. This paper also studies the strategic management and operation management of real-time risk control. Through the SWOT analysis of real-time risk control, the advantages and disadvantages of P Company's real-time risk control system are analyzed. Opportunities and threats. By using the 4P method of Internet operation, this paper studies the establishment of real-time risk control operation system and the construction of operation team in P Company. Finally, through the analysis of the effect of the implementation of real-time risk control in P Company, The actual risk results of real-time risk control are described, and the shortcomings of the new system and suggestions for improvement are analyzed.
【学位授予单位】：东华大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09;TP273

【相似文献】