基于集成学习的P2P网贷平台风险识别研究
首发时间:2018-11-09
摘要:从P2P平台的总体安全性考虑,实现了对P2P平台风险的有效甄别,并利用Random Uniform Forests(rUF)算法对影响P2P平台风险识别的关键因素深入分析。研究发现:相较于单一的传统分类算法,集成学习对甄别P2P平台风险更加有效;平台平均利率波动、资金净流入、未来待还金额波动、平均借款期限、正常运营时间、平台实力等信息对识别P2P平台风险具有重要作用;平台实力、未来待还金额波动由于与其他变量的强相互作用而间接影响平台风险;监管政策的频繁变动一定程度上会对平台运营造成压力,但同时也促进平台的优胜劣汰。rUF算法的变量重要性度量方法能够细致分析各指标对P2P平台风险识别的影响大小和方式,有助于对平台风险进行监控和管理。
For information in English, please click here
The Risk Identification of P2P Platforms based on Ensemble Learning
Abstract:In this paper, the overall security of the P2P platforms is considered seriously, different classification models are built to identify the risk of P2P platforms, and degree of importance of various factors which influence the risk identification of P2P platforms is debated based on Random Uniform Forests. Results show that ensemble learning are more effective in risk identification compared to the single traditional classification algorithm; Some indicators such as Average Rate Volatility, Net Capital Inflow, Repay Receivable Volatility, Average Loan Term, normal Operation Time, Platform strength play important roles in risk identification of P2P platforms; Platform strength and Repay Receivable Volatility have indirect impacts on the risk of platforms due to strong interaction with other indicators; Frequent changes of regulatory policies will put pressure on the operation of P2P platforms to some extent, but at the same time, it will also promote the survival of the fittest of P2P platforms. The variable importance measurement method of rUF algorithm can analyze the impact size and mode of each indicator on P2P platform risk identification and help to monitor and manage platform risk.
Keywords: Peer to Peer lending Risk analysis Ensemble Learning the Factors of Risk Identification
基金:
引用
No.****
同行评议
勘误表
基于集成学习的P2P网贷平台风险识别研究
评论
全部评论0/1000