摘要
个人信用评价问题研究中,需要建立较多的虚拟变量作为解释变量.Group Lasso可以将相关的虚拟变量作为组进行整体剔除或保留在模型中.结合具体的个人信贷数据,应用Group Lasso方法进行变量选择建立Logistic模型,并与全模型、向前选择和向后选择建立的Logistic模型进行比较,发现Group Lasso方法建立的模型,在变量解释和预测正确率上,都是最优的.
Dummy variables would be established in the analysis of personal credit evaluation.The group lasso executes variable selection on group variables,so the related dummy variables could be groupwise eliminated or reserved in the model.The paper fitted Logistic model based on group lasso using the personal credit data,and compared with the methods of full model,forward selection and backward selection.The results show that the group lasso is optimal in variable explanation and forecast accuracy.
出处
《数学的实践与认识》
北大核心
2015年第6期89-94,共6页
Mathematics in Practice and Theory
基金
国家社会科学基金项目(13BTJ004)
