融入软信息的P2P网络借贷违约预测方法

发布时间：2018-07-03 05:44

本文选题：PP借贷 + 违约预测　；参考：《中国管理科学》2017年11期

【摘要】：在P2P网络借贷中,预测借款的违约概率是用户信用评价的关键,也是借贷平台与投资者关注的重点问题。由于P2P平台所获取的用户财务信息有限,P2P借款信用评价和违约预测面临新的挑战。本文结合P2P平台的信息特点,提出一种融入软信息的网络借款违约预测方法。首先利用主题模型抽取并量化文本软信息中的相关变量,进而分析不同软信息变量对借款违约的影响关系;其次,设计了一种两阶段的变量选择方法对软硬信息进行组合筛选;最后,引入随机森林算法构建融入软信息的违约预测模型,并结合P2P平台的真实数据进行实证分析。结果表明,在P2P借款的违约预测模型中融入有价值的软信息可以提高预测准确率。
[Abstract]:In P2P network lending, predicting the default probability of loan is the key of user credit evaluation, and also the key issue of loan platform and investors. Due to the limited financial information obtained by P2P platform, P2P loan credit evaluation and default prediction face new challenges. According to the information characteristics of P2P platform, this paper proposes a network loan default prediction method which integrates soft information. Firstly, we use the topic model to extract and quantify the relevant variables in the text soft information, and then analyze the influence of different soft information variables on the loan default. Secondly, a two-stage variable selection method is designed to select the soft and hard information. Finally, the stochastic forest algorithm is introduced to construct the default prediction model with soft information, and the real data of P2P platform are analyzed empirically. The results show that the prediction accuracy can be improved by incorporating valuable soft information into the default prediction model of P2P loan.
【作者单位】：合肥工业大学管理学院;
【基金】：国家自然科学基金资助项目(71731005,71571059)
【分类号】：F724.6;F832.4

【相似文献】