| 基于主成分的稀疏贝叶斯信用分类研究 |
| |
| 引用本文： | 盛静文,于艳丽,江开忠.基于主成分的稀疏贝叶斯信用分类研究[J].教育技术导刊,2009,19(8):113-116. |
| |
| 作者姓名： | 盛静文 于艳丽 江开忠 |
| |
| 作者单位： | 上海工程技术大学 数理与统计学院，上海 201620 |
| |
| 摘 要： | 针对传统信用评价方法分类精度较低、数据集属性变量间存在相关性等问题，提出基于主成分分析的稀疏贝叶斯学习（PCA-SBL）算法。首先对数据集特征变量进行主成分分析，使降维后的变量无相关性|其次，对主成分分析后的数据进行稀疏贝叶斯分类|最后将 PCA-SBL 分类方法分类精度与传统分类方法精度进行比较。分析发现，在 German Credit Data 和 Australian Credit Data 上，与传统 KNN、朴素贝叶斯、SVM、随机森林、决策树相比，改进的 SBL 算法分类精度平均提高了 5.26%、4.65%、2.11%、2.125%、4.66%，与稀疏贝叶斯学习算法（SBL）相比，平均提高 0.965%，从而证明 PCA-SBL 算法具有更高的分类效果。
|
| 关 键 词： | 信用评价 主成分分析 稀疏贝叶斯学习 信用分类 |
| 收稿时间： | 2019-11-18 |
Research on Sparse Bayesian Credit Classification Based on Principal Component |
| |
| SHENG Jing-wen,YU Yan-li,JINANG Kai-zhong.Research on Sparse Bayesian Credit Classification Based on Principal Component[J].Introduction of Educational Technology,2009,19(8):113-116. |
| |
| Authors: | SHENG Jing-wen YU Yan-li JINANG Kai-zhong |
| |
| Institution: | School of Mathematics，Physics and Statistics，Shanghai University of Engineering Science，Shanghai 201620，China |
| |
| Abstract: | Aiming at the low classification accuracy of traditional credit evaluation methods and the correlation between data set attribute variables，this paper proposes a sparse Bayesian learning algorithm based on principal component analysis（PCA-SBL）. Firstly，the principal component analysis of the characteristic variables of the data set is carried out，so that the variables have no correlation after dimensionality reduction. Secondly，the sparse Bayesian classification is performed on the data after principal component analysis. Finally，by comparing the classification accuracy of the PCA-SBL classification method with the accuracy of the traditional classification method，it is found that the improved SBL increases the classification accuracy by 5.26%，4.65%，2.11%，2.125% and 4.66% averagely when compared with the traditional K-Nearest Neighbour（KNN），Naive Bayes，support vector machine，random forest and decision tree respectively on real-world German and Australian credit datasets. It also improves 0.965% averagely when compared with sparse Bayesian learning（SBL）algorithm. This proves that the proposed PCA-SBL algorithm has a higher classification effect. |
| |
| Keywords: | credit risk evaluation principal component analysis sparse Bayesian learning credit classification |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
