大数据背景下网络借贷平台用户二次贷款预测研究

发布时间：2018-03-25 17:00

本文选题：网络借贷　切入点：精准借贷　出处：《山西财经大学》2017年硕士论文

【摘要】：网络借贷是我国新型金融体系建设的重要组成部分,有利于解决中小企业融资难的问题,一定程度上促进了我国经济的发展。然而不少网络借贷平台面临生存危机,原因是平台自身经营成本高,没有制定科学的营销策略,投入大量线下和线上推广却难见成效。为了能够解决上述问题,网络借贷平台需要借助大数据技术对用户行为进行分析和预测,从而更加了解客户需求,制定有针对性的营销策略。网络借贷平台二次贷款预测研究有利于网络借贷平台对用户是否由再次贷款需求进行预测,从而提高营销转化率,进而提高其盈利能力。本文以网络借贷平台二次贷款预测为目标,研究了如下内容:首先结合网络借贷和精准营销理论,提出了精准借贷这一概念,并分析了大数据背景下网络借贷平台精准借贷的流程。其次,对网络借贷平台的用户数据进行分析和整理,初步构建了网络借贷平台用户二次贷款指标体系;利用具有较好的特征选择能力的弹性网方法对指标进行筛选,选择出具有重要意义的指标。然后,在此基础上将稀疏贝叶斯网络结构学习方法应用于网络借贷用户行为分析中,并证明该方法优于一般的贝叶斯网络;通过分析指标之间的关系,研究出网络借贷平台用户的行为关系和重要特点。接着,充分利用Xg-boost这一集成算法的优势,将其应用于网络借贷用户二次贷款预测中,并采用分类正确率和时间复杂度两个评价指标进行预测模型评估,取得良好的效果,证明该算法适合在类似的问题中被广泛推广和应用。最后根据分析结果并结合网络借贷发展形势,提出平台为了实现精准借贷可以充分利用用户信息,构建用户行为网络和预测模型,为制定精准借贷策略提供数据支持。本文得出的主要结论有:(1)网络借贷平台用户二次贷款预测为平台精准借贷提供数据支持,对于平台服务策略制定具有指导性意义;(2)弹性网方法在变量筛选中的应用有选择重要变量构建模型,提高模型分类性能;(3)消费记录和社会资本的引入可以提高模型的分类正确率;(4)Xg-boost模型在网络借贷平台二次贷款的预测性能中优于其他集成分类器且时间复杂度低,适宜推广。
[Abstract]:Network lending is an important part of the construction of new financial system in China, which is helpful to solve the problem of financing difficulties for small and medium-sized enterprises, and to some extent promotes the development of our economy. However, many online lending platforms are faced with survival crisis. The reason is that the platform itself has high operating costs, no scientific marketing strategy, a large amount of offline and online promotion is difficult to see results. Online lending platforms need to use big data technology to analyze and predict user behavior in order to better understand customer needs. The research on the secondary loan forecast of the network loan platform is helpful for the network loan platform to predict whether the users' demand for the second loan will be predicted by the second loan, thus increasing the marketing conversion rate. In order to improve its profitability, this paper aims at the second loan forecast of network lending platform, and studies the following contents: firstly, combining the theory of network lending and precision marketing, the concept of precision lending is put forward. And analyzes the network lending platform precision lending process under the background of big data. Secondly, the user data of the network lending platform is analyzed and collated, and the user secondary loan index system of the network lending platform is preliminarily constructed. Using the elastic network method with better feature selection ability to screen the index and select the important index. Then, the sparse Bayesian network structure learning method is applied to the behavior analysis of the network lending user, based on which the sparse Bayesian network structure learning method is applied to the behavior analysis of the network loan user. It is proved that this method is superior to the general Bayesian network, and through the analysis of the relationship between the indexes, the behavior relationship and important characteristics of the users of the network lending platform are studied. Then, the advantages of the Xg-boost integration algorithm are fully utilized. It is applied to the secondary loan prediction of network loan users, and two evaluation indexes, the classification accuracy and the time complexity, are used to evaluate the prediction model, and good results are obtained. It is proved that the algorithm is suitable to be widely used in similar problems. Finally, according to the analysis results and combined with the development situation of network lending, the platform can make full use of user information in order to realize precision lending. Build user behavior network and forecast model to provide data support for the development of precision lending strategy. The main conclusions of this paper are: 1) the network lending platform user second loan forecast provides data support for the platform precision lending. For the application of the flexible network method in variable selection, the important variables are selected to build the model, which is of guiding significance to the development of platform service strategy. To improve the classification performance of the model, the consumption record and the introduction of social capital can improve the classification accuracy of the model and the 4Xg-boost model is superior to other integrated classifiers in the prediction performance of the secondary loan in the network lending platform, and the time complexity is low, and the model is suitable for popularization.
【学位授予单位】：山西财经大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F724.6;F832.4

【参考文献】