基于GBDT模型的P2P网贷平台运营风险预警研究

Operation Risk Early Warning of P2P Online Lending Platform Based on GBDT Model

近年来,大量P2P网贷平台倒闭和跑路,严重损害了投资者利益,进而影响行业发展,科学预警P2P网贷平台风险对促进网贷行业健康发展有重要意义。基于358个网贷平台的样本数据,从平台实力、平台风控能力、平台运营能力、平台治理水平、平台合规性、标的特性与收益率等六个维度,构建P2P网贷平台运营风险预警指标体系,以梯度提升决策树(GBDT)模型为基础构建P2P网贷平台运营风险预警模型。结果表明:与逻辑回归和决策树等模型相比,GBDT模型对网贷平台风险有更好的预测性能。

In recent years, a large number of P2 P online lending platforms have been closed down and run away, which has seriously damaged the investors’ benefit. How to make a scientific early warning of the risk of P2 P platform is of great significance to promote the healthy development of P2 P online lending industry. Based on the data of 358 sample platforms, the paper builds a warning index system of P2 P online lending platform operation risk, the system includes the following indicators such as platform strength, platform risk control capabilities, platform operation capabilities, platform governance levels, platform compliance, and loaning’s characteristics and yield. On this basis, the paper constructs an early warning model for the operation risk based on gradient boosting decision tree(referred to as the GBDT) model. The results show that compared with logistic regression, decision tree and other models, the GBDT model has better prediction performance.