基于bagging方法的企业电子商务信用风险预警研究

发布时间：2018-04-17 02:37

本文选题：电子商务信用 + 风险预警　；参考：《广西大学》2017年硕士论文

【摘要】：近些年,随着网络的快速发展,人们开始利用网络开展新的商务活动模式——电子商务。在电子商务飞速发展的过程中,制约其发展的一些基础性问题(例如:交易支付、物流配送等)得到了逐步的解决,但是,电子商务活动中的欺诈事件依旧很多,诚信缺失问题依然严重,信用风险已经凸显为电子商务健康快速发展的瓶颈。本文立足于电子商务迅速发展但是又受制于信用风险的现状,将企业电子商务信用风险预警系统进行优化,优化后的系统可以及时准确的发布预警信息,尽量减少企业的损失,同时也为有关信用管理部门提供参考信息,使其更好的监管企业。电子商务交易双方的企业可以利用该系统评测自己、交易对象甚至竞争对手的信用状况,以此来指导后续的决策。本文的重点在风险信息收集和风险识别两个方面。在风险信息收集方面,本文针对以往相关文献里的指标体系进行考察,不但保留了部分财务指标,还创新性的引入了“网络交易纠纷率”和“研发投入占营业收入比例”两个指标来更好的反应服务能力和发展能力两部分的内容,使企业电子商务信用风险得到更加全面的反映。在风险识别方面,本文分析了机器学习里的支持向量机方法和集成学习方法在本文的可行性,创新性的将集成学习方法引入企业电子商务信用风险预警领域内,选择了 bagging算法来优化支持向量机,获得了很好的分类效果。经过优化的企业电子商务信用风险预警系统的风险识别准确率更高,说明加入新的预警指标的指标体系是合理的,引入集成学习方法在企业电子商务信用风险预警领域内是可行的。由于机器学习方法有自学习性,本文优化后的预警系统具一定的实用性。同时,本文选择将警情分为三级的方式,简单明了,所以本文整合的预警系统具有易推广的特点。
[Abstract]:In recent years, with the rapid development of the network, people begin to use the network to develop a new business model-e-commerce.In the process of the rapid development of electronic commerce, some basic problems (such as transaction payment, logistics distribution, etc.) which restrict the development of electronic commerce have been gradually solved. However, there are still a lot of fraud incidents in electronic commerce activities.Lack of integrity is still a serious problem, credit risk has been highlighted for the rapid and healthy development of e-commerce bottleneck.In this paper, based on the rapid development of electronic commerce but also subject to the status quo of credit risk, the enterprise e-commerce credit risk warning system optimization, the optimized system can timely and accurate release of early warning information, as far as possible to reduce the losses of enterprises.At the same time also for the relevant credit management departments to provide reference information for better supervision of enterprises.The system can be used by enterprises of both sides to evaluate the credit status of themselves, trading objects and even competitors, so as to guide the subsequent decisions.This paper focuses on risk information collection and risk identification.In the aspect of risk information collection, this paper investigates the index system in the previous relevant literature, and not only retains some financial indicators,It also innovatively introduces the two indicators of "online transaction dispute rate" and "R & D investment as a proportion of operating income" to better respond to both service and development capabilities.Make the enterprise e-commerce credit risk more comprehensive reflection.In the aspect of risk identification, this paper analyzes the feasibility of support vector machine (SVM) and integrated learning in machine learning, and innovatively introduces the integrated learning method into the early warning field of enterprise electronic commerce credit risk.Bagging algorithm is chosen to optimize support vector machine (SVM), and good classification effect is obtained.The risk identification accuracy of the optimized e-business credit risk early warning system is higher, which shows that it is reasonable to add the new early warning index index system.It is feasible to introduce the integrated learning method in the field of enterprise e-commerce credit risk warning.Because the machine learning method has the self-study habit, the optimized early warning system in this paper has certain practicability.At the same time, this paper chooses to divide the police situation into three levels, which is simple and clear, so the integrated early warning system is easy to be popularized.
【学位授予单位】：广西大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F275

【参考文献】