基于改进朴素贝叶斯模型的信用评估

Credit Evaluation Based on Improved Naive Bayesian Model

摘要: 伴随着消费信贷的快速发展，激起了对个人信用评估的需求。为了帮助金融机构更好的了解个人信用情况，结合快速独立分量分析方法(FastICA)和线性判别分析(LDA)提取数据特征的优势，提出了一种基于改进朴素贝叶斯分类算法的信用评估模型——FastICA-LDA-NB。将该模型应用于UCI上的德国个人信用数据集，在精确率、召回率两个评价指标值上表明所提模型具有较好的信用评估效果。

Abstract: Along with the rapid development of consumer credit, the demand for personal credit assessment has been aroused. To help financial institutions better understand their personal credit situation, combining the advantages of Fast Independent Component Analysis method (FastICA) and Linear Discriminant Analysis (LDA) to extract data features, a credit evaluation model called FastI-CA-LDA-NB is proposed, which is based on the improved Naive Bayesian classification algorithm. Applying the model to the UCI German personal credit data set, the proposed model has a good credit evaluation effect on the two evaluation index values of accuracy rate and recall rate.