基于GA-CFS属性选择的个人信用评估模型

摘要

Attribution selection method could reduce data redundancy and the data dimension degree effectively and efficiently.This paper applies attribution selection of GA-CFS method to personal credit evaluation, and uses by the heuristic “merit” as GA fitness function to optimize personal credit index system through constructing a personal credit evaluation model based on attribution selection of GA-CFS.In addition, we compares with ID3, NB, Logistic, SMO and combination of GA-CFS attribute selection methods and the four classification algorithms in Australian data sets.Experiment results show that this model not only reduces the dimension of personal credit index and the amount of training data but also has higher classification accuracy than the personal credit evaluation model based on single classifier.%属性选择可以有效减少数据的冗余度和降低数据的维度,将GA-CFS属性选择方法引入个人信用评估中,利用CFS评价得到的启发式"价值"作为GA的适应度函数来对个人信用指标体系优化,建立了基于GA-CFS属性选择的个人信用评估模型.在Australian数据集上比较了ID3、NB、Logistic、SMO与GA-CFS属性选择方法和这四种分类算法分别结合执行的结果.实验结果表明,基于GA-CFS属性选择的个人信用评估模型降低了个人信用指标的维度,减少了学习所需的数据量,而且比基于单分类器的个人信用评估模型具有更高的分类准确率.