基于SVM_AdaBoost模型的上市公司退市预警研究

发布时间：2018-06-19 02:39

本文选题：退市预警 + 支持向量机　；参考：《华南理工大学》2013年硕士论文

【摘要】：退市制度是资本市场整体框架构成的重要组成部分，一个健康合理的资本市场既要保证经营业绩好的企业能进场，又要保证经营效益差的企业被清理出场。我国已于2012年相继颁布了创业板退市新规和主板与中小板退市新规，这标志着我国证券市场多年来上市公司有进无退历史的终结。退市风险存在于我国沪深两市中的部分上市公司之中，尤其是ST标志上市公司。对于上市公司风险识别和处置是保证公司有效运行的核心内容，建立有价值的上市公司退市风险预警模型，尽早识别上市公司是否有退市风险，有利于做到风险的事前控制，这是保证投资者合法权益，，降低市场风险的有效途径。本文使用SVM_AdaBoost强分类器模型构建上市公司退市预警模型。支持向量机（SVM）是数据挖掘中的新方法，AdaBoost算法作为一种通用的学习算法，可以提高任一给定算法的性能。使用AdaBoost算法连接若干个不同核函数的SVM，可以得到分类精度更高的强分类器SVM_AdaBoost模型。本文选取200家上市公司作为样本，先粗选17个指标，后使用独立样本T检验精选9个指标作为预警指标。然后对9个指标进行归一化处理，消除量纲差异的影响，最终使用AdaBoost算法构建基于径向基核函数和多项式核函数的10个不同的SVM的SVM_AdaBoost强分类器，进行退市预警。研究发现：相对单一SVM，SVM_AdaBoost对70家测试样本公司的分类性能由92.8571%提高到95.7143%，这显示了SVM_AdaBoost强分类器模型有较在退市预警研究中有较好的应用价值。
[Abstract]:The delisting system is an important part of the overall framework of the capital market. A healthy and reasonable capital market should not only guarantee the entry of enterprises with good operating performance, but also ensure that enterprises with poor operating efficiency are cleared out. In 2012, China has promulgated the new rules for delisting of gem and the new rules for delisting of main board and small board, which marks the end of the history of listed companies in the stock market of our country for many years. Delisting risk exists in some listed companies in Shanghai and Shenzhen stock markets, especially St mark listed companies. Risk identification and disposal of listed companies is the core content to ensure the effective operation of the company. Establishing a valuable early warning model of delisting risks of listed companies and identifying whether there are delisting risks of listed companies as soon as possible is beneficial to the prior control of risks. This is an effective way to ensure the legitimate rights and interests of investors and reduce market risks. In this paper, SVMAdaBoost strong classifier model is used to construct the delisting warning model of listed companies. Support Vector Machine (SVM) is a new method in data mining. As a general learning algorithm, AdaBoost algorithm can improve the performance of any given algorithm. Using AdaBoost algorithm to connect several SVMs with different kernels, a stronger SVMStackAdaBoost model with higher classification accuracy can be obtained. In this paper, 200 listed companies are selected as samples, 17 indexes are selected first, and then 9 indexes selected by independent sample T test are used as early warning indexes. Then the nine indexes are normalized to eliminate the influence of dimensional difference. Finally, the SVMAdaBoost strong classifier of 10 different SVM based on radial basis function and polynomial kernel function is constructed using AdaBoost algorithm to carry out delisting warning. It is found that the classification performance of SVM _ S _ AdaBoost is improved from 92.8571% to 95.7143%, which shows that SVM _ AdaBoost strong classifier model has better application value in delisting and early warning research.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.51;F224

【参考文献】