中文摘要：

本文以大数据征信为视角,认为空间维度上借款人多角度、多层次信息的交叉复现,以及时间维度上借款人社会活动信息的持续呈现能够更加准确地反映借款人信用状况,进而构建了基于贝叶斯网络（Bayesian network,BN）的P2P借款人信用评价模型。研究表明,贝叶斯网络为P2P借款人多维信息间的复杂关系提供了统一的表达方式;基于贝叶斯网络推理的样本内信用评价准确率高达87%,提高信用评价的概率值临界点能够显著增强信用评价模型的精准性;样本外信用评价准确率超过90%,增加训练数据集能明显提高信用评价模型的精准性;通过对比不同信息维度模型的评价准确率,也可验证所构建的P2P借款人信用评价模型是稳健的。

英文摘要：

From the perspective of big data inquiry,this paper introduces an appraisal model for P2 P credit ratings. This model uses a multi-angle,multi-level model that places credit information in a spatial dimension to cross retrieve borrowers＇ data reflecting credit history continuously. Based on this concept,a Bayesian network was used to test the effectiveness of the P2 P borrower credit evaluation model and attempted to solve some information uncertainty issues. To test the model empirically,this model was performed on the data obtained from Prosper Lending Company. Some results of the Bayesian network include an unified expression for the complicated relation between P2 P borrowers; the inference credit evaluation accuracy reached 87%within the samples based on the Bayesian network; the critical probability of credit evaluation value increased; the inference credit evaluation was accurate for more than 90% of the samples; increasing the data set could significantly improve the effectiveness of the credit evaluation model; better accuracy and efficiency in evaluating the borrower＇s credit status can be achieved if the dimension of the borrower＇s information is greater. By comparing the accuracy of the different dimensions＇ models,the evaluation model that verifies the P2 P borrower＇s shows a strong robustness.