SCAD惩罚逻辑回归的财务预警模型  

Financial Early-Warning Model Using Logistic Regression with SCAD Penalty

作　　者：刘遵雄[1] 黄志强[1] 孙清[2] 张恒[1] 

机构地区：[1]华东交通大学信息工程学院,江西南昌330013 [2]西安财经学院信息学院,陕西西安710100

出　　处：《统计与信息论坛》2012年第12期21-26,共6页Journal of Statistics and Information

基　　金：国家自然科学基金项目<高维数据非负稀疏特征抽取及聚类算法研究>(61065003);教育部人文社会科学研究规划基金项目<融合变量选择的上市公司财务预警模型研究>(10YJC630379);江西省自然科学基金项目<文本聚类中非负稀疏特征抽取算法研究>(2010GZS0034)

摘　　要：作为一种有监督学习算法,逻辑回归(Logistic Regression,LR)已广泛应用于财务危机建模分析,但其潜在地存在过拟合问题。鉴此,提出一种基于平滑削边绝对偏离(Smoothly Clipped Absolute Deviation,SCAD)惩罚逻辑回归的财务预警模型。该模型不仅能很好地解决模型过拟合问题,而且还可以同时实现变量选择和模型系数估计,并提高了模型的解释性。结合沪深股市A股制造业上市公司的财务数据进行实证研究,同时对比一般的L1正则化和L2正则化逻辑回归模型的预警效果进行实证分析,实验结果表明:SCAD惩罚逻辑回归模型具有较好的分类效果和较强的经济解释能力。Logistic regression （LR） is a supervised learning algorithm. And it is widely used in financial early-warning models and analysis. But its model has the potential over-fitting problem. Aiming at this problem, the financial early-warning model using LR with smoothly clipped absolute deviation （SCAD） penalty is put forward. The model can well solve the over-fitting problem and perform variable selection and model parameter estimation simultaneously. And the model＇s explanatory power is improved. Experiments are implemented on the financial data of A-share manufacturing listed companies of the Shanghai and Shenzhen stock markets. Compared with general, L1 norm penalized and L2 norm penalized LR model, the LR model with SCAD penalty has better classification results and stronger economic interpretations.

关 键 词：财务预警 逻辑回归 SCAD惩罚 

分 类 号：C811[社会学—统计学]