摘要
利用随机森林特征选择算法,对信用评估的可用指标集进行特征选择,在此基础上建立基于随机森林融合朴素贝叶斯的信用评估模型.选取UCI数据库中的German数据集进行实证研究,结果表明,通过随机森林进行特征选择的随机森林融合朴素贝叶斯模型具有更高的预测准确度.
A new credit assessment model is applied based on random Forest algorithm to calculate the importance of individual character.This model provides the basis for the selection in its evaluation index system by random Forest.The experimental results on German dataset show that the credit assessment model based on random Forest and naive Bayes has higher prediction accuracy.
作者
叶晓枫
鲁亚会
YE Xiao-feng LU Ya-hui(School of Mathematics and Information Science, North China University of Water Resources and Electric Power, Zhengzhou 450046, Chin)
出处
《数学的实践与认识》
北大核心
2017年第2期68-73,共6页
Mathematics in Practice and Theory
基金
河南省基础与前沿技术研究项目(142300410401)
关键词
随机森林
朴素贝叶斯
特征选择
信用评估
random forest
naive bayes
feature selection
credit assessment
