网贷平台数据的随机森林预测模型实证分析

Empirical Analysis of Random Forest Prediction Model for Online Loan Platform Data

针对P2P网络借款平台存在着借款人与投资人信息的不对称性、借款成功率低下这些阻碍网贷平台未来发展的问题,依托已有的非均衡少量样本数据,构建随机森林预测模型并进行实证分析.结果表明,随机森林模型对各级风险样本的正确识别率接近100%,说明该模型具有很高的实用价值和很好的预测能力.与其他模型相比,随机森林比线性模型和广义线性模型的预测精度高,抗噪声性能良好,且能够有效估计模型中各变量在分类中的重要性.通过对各评价指标的重要程度及其与借贷成功率之间的关系的分析,可为P2P网贷发展提供理论依据和实践基础.

Aiming at the problems that hinder the future development of online lending platforms, such as asymmetry of borrower and investor information and low success rate of borrowing for P2 P network borrowing platform, a random forest prediction model was constructed and empirically verified based on the existing unbalanced small sample data. The results show that the random forest model has a correct rate of recognition of risk samples at all levels close to 100%, indicating that the model has high practical value and good predictive ability. Compared with other models, random forests have the advantages of higher accuracy than linear models and generalized linear models, and have good anti-noise performance, which can effectively estimate which variables in the model are more important in classification. Finally, the importance of each evaluation index and its relationship with the success rate of borrowing are analyzed, which can provide theoretical and practical basis for P2 P online loan development.