摘要
机器算法中存在许多不同类型和方式的运行模式,而在诸多算法之中,集成学习的算法是一种基于统计理论以计算机实现的良好机器学习方法.阐述了集成学习的基本思想和实现步骤,运用Bagging集成学习算法试图建立一个个人信用评估模型,以期取得更好的预测结果.运用信息增益法筛选指标,采用V折交叉确认法,利用UCI的信用数据对单个分类器、Bagging集成分类器模型的分类精度和稳健性进行试验比较.结果表明,Bagging-决策树有效的提高了样本的精确性,在个人信用评估的分析中占有较强的优势.
Ensemble learning algorithm is a good machine learning method based on statistical theory with computer to reaiize.This paper introduces the basic thought and realized steps of ensemble learning algorithms,we use an improved integrated classifier,Bagging,to build a model for personal credit scoring,hoping to obtain better results.The classification accuracy and the robustness of the model are compared with single classifier and bagging using the UCI datasets.The result shows that Bagging-decision tree can effectively improve the classification accuracy and have stronger advantage for the evaluation of personal credit.
出处
《数学的实践与认识》
北大核心
2016年第8期90-98,共9页
Mathematics in Practice and Theory
关键词
个人信用评估
集成算法
BAGGING
信息增益
personal credit evaluation
ensembie learning algorithm
bagging
information Gain
