不同机器学习方法在互联网征信的表现
首发时间:2019-04-30
摘要:本文基于拍拍贷网提供的互联网用户大数据,分析用户的信用表现预测用户是否会贷款违约。本文一共使用了8种不同的机器学习方法,针对每种方法,比较预测结果与实际结果发现,XGBoost预测错误率仅有13.3%,而random forest预测错误率有33%,Adaptive boosting预测错误率有38%。因此在本文中相对出众的机器学习方法是XGBoost。
关键词: 互联网征信 XGBoost Adaptive boosting
For information in English, please click here
The performance of different machine learning methods in the Internet for credit reporting
Abstract:This article is based on the big data of Internet users provided by the pat network, and analyzes the credit performance of users to predict whether users will default on loans. In this paper, eight different machine learning methods are used. For each method, comparing the prediction results with the actual results, the XGBoost prediction error rate is only 13.3%, while the random forest prediction error rate is 33%. The Adaptive boosting prediction error rate has 38%. So the relatively superior machine learning method in this article is XGBoost.
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
不同机器学习方法在互联网征信的表现
评论
全部评论0/1000