兰州理工大学学报

兰州理工大学学报 ›› 2021, Vol. 47 ›› Issue (6): 100-107.

• 自动化技术与计算机技术 • 上一篇    下一篇

基于Adaboost算法结合DEGWO-SVM的财务困境预测

朱昶胜*1,田慧星1,冯文芳2   

  1. 1.兰州理工大学 计算机与通信学院, 甘肃 兰州 730050;
    2.兰州理工大学 经济管理学院, 甘肃 兰州 730050
  • 出版日期:2021-12-28 发布日期:2021-12-28
  • 通讯作者: 朱昶胜(1972-),男,甘肃秦安人,博士,教授,博导.Email:zhucs_2008@163.com
  • 基金资助:
    国家自然科学基金(72161026),甘肃省社科规划项目(19YB141)

Financial distress prediction based on Adaboost algorithm combined with DEGWO-SVM

ZHU Chang-sheng1, TIAN Hui-xing1, FENG Wen-fang2   

  1. 1. College of Computer and Communication, Lanzhou Univ. of Tech., Lanzhou 730050, China;
    2. College of Economics and Management, Lanzhou Univ. of Tech., Lanzhou 730050, China
  • Online:2021-12-28 Published:2021-12-28

PDF

284

摘要： 针对支持向量机(SVM)在企业财务困境预测研究中存在参数选择困难、分类准确率低的问题,提出了一种新的Adaboost-DEGWO-SVM组合模型.首先,通过对2017年全部A股上市公司的财务数据进行数据预处理,提取1∶1的困境公司(ST)和正常公司组成建模数据集;然后,利用差分进化算法(DE)改进灰狼优化算法(GWO)来提高其全局搜索能力,以解决灰狼算法易陷入局部最优的问题,从而实现对SVM参数c和γ的寻优;最后,通过Adaboost算法提高了DEGWO-SVM模型的分类能力.实验结果表明,Adaboost-DEGWO-SVM组合预测模型具有明显的困境预测优势,与DEGWO-SVM相比,分类准确率提高了4.34%,Ⅰ类错误和Ⅱ类错误分别降低了0.043 5;与单一SVM相比,分类准确率提高了13.04%,Ⅰ类错误、Ⅱ类错误分别降低了0.130 4、0.130 5,是一种潜在的企业财务困境预测方法.

关键词: 困境预测, 支持向量机, 改进的灰狼优化算法, Adaboost算法

Abstract: A new Adaboost-DEGWO-SVM combination model is proposed in this paper to solve the problems of difficult parameter selection and low classification accuracy of support vector machine (SVM) in financial distress prediction. Firstly, the financial data of all A-share listed companies were preprocessed in 2017, and 1∶1 modeling dataset of distressed companies (ST) and normal companies was extracted. Then, the difference evolutionary algorithm (DE) was used to improve the global search capability of GWO, and to solve the problem that the gray wolf algorithm is prone to fall into the local optimal, so as to realize the optimization of the parameters of SVM. Finally, Adaboost algorithm is used to improve the classification ability of DEGWO-SVM model. Experimental results show that the proposed Adaboost-DEGWO-SVM combined forecasting model has obvious advantage in dilemmaprediction, compared with DEGWO-SVM, the classification accuracy rate isincreased 4.34%, Ⅰ errors and Ⅱ errors aredecreased by 0.043 5, and compared with the single SVM, theclassification accuracy rate is increased 13.04%, Ⅰ errors and Ⅱ errors decreased 0.130 4 and 0.130 5, it is a kind of potential corporate financial distress prediction model.

Key words: distress prediction, support vector machine, improved grey wolf optimization algorithm, Adaboost algorithm

中图分类号: