基于支持向量机的制造业上市公司财务预警分析

发布时间：2018-02-23 17:55

本文关键词： 财务危机支持向量机上市公司加权平均　出处：《长沙理工大学》2013年硕士论文　论文类型：学位论文

【摘要】：上市公司是我国资本市场发展的基础,其运行的好坏直接关系到资本市场的兴衰,因此我们需要在上市公司发生财务危机之前,对其财务状况做预警分析。本文以支持向量机为研究工具,旨在建立有效的财务预警模型。在综合了国内外研究现状基础上结合对财务危机概念的界定,利用支持向量机理论基础建立数学模型进行财务危机的预测。本文选取2008-2009年的20家发生财务危机的上市公司和与其配对的20家财务正常的公司为建模样本,以2010年的30家上市公司(其中财务正常和发生财务危机的上市公司各15家)为验模样本,用来检验模型的好坏。一共选取了19个财务指标,通过一系列的指标筛选,利用上市公司发生财务危机的前三年的财务指标数据分别建立模型,来预测验模样本是否会发生财务危机,预测的精度分别为63.3%,90%,90%。由于只是运用单年的财务指标数据来建立模型,数据的信息量比较小,不同年份的财务指标数据产生的预测精度也不同,因此本文试图将上市公司发生财务危机的前三年的财务指标数据做加权平均,权重为前三年财务指标单年的预测精度与三年总的预测精度之比,这样既能保证更大范围的运用数据信息,又能根据不同年份对预测精度的贡献程度不同选择不同的权重,使得最终的预测精度更为准确。经过试验,得到的预测精度为96.7%,这比使用单年财务指标数据进行预测的精度要高。本文以制造业的上市公司为样本进行试验得到了较好的结果,可以尝试将此种方法推广到其他行业的财务危机预警中。
[Abstract]:The listed company is the foundation of the capital market development in our country, and its operation is directly related to the rise and fall of the capital market. In this paper, the support vector machine is used as the research tool to establish an effective financial early warning model. On the basis of synthesizing the present research situation at home and abroad, combining with the definition of the concept of financial crisis, using the support vector machine theory to establish the mathematical model to predict the financial crisis. In this paper, 20 listed companies with financial crisis and 20 companies with normal financial situation in 2008-2009 were selected as modeling samples. In 2010, 30 listed companies (including 15 listed companies with normal financial affairs and 15 listed companies with financial crisis) were used to test the model. A total of 19 financial indicators were selected and selected through a series of indicators. Using the financial index data of the first three years of the financial crisis of a listed company to establish models separately to predict whether there would have been a financial crisis or not, the accuracy of the prediction is 63.33 / 90 / 900.Because the model is only established by using the single year's financial index data, The amount of information in the data is relatively small, and the forecasting accuracy of the financial index data in different years is different, so this paper tries to make the weighted average of the financial index data of the first three years of the financial crisis of listed companies. The weight is the ratio of the forecasting precision of the first three years to the total forecast precision of three years, which can not only guarantee the use of data information in a wider range, but also select different weights according to the contribution of different years to the forecast accuracy. The accuracy of the final prediction is more accurate. After the experiment, the prediction accuracy is 96. 7, which is higher than the prediction accuracy using the single year financial index data. This paper takes the listed companies of manufacturing industry as the sample to carry out the experiment and gets better results. Try to extend this approach to financial distress warnings in other industries.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F275;F832.51;F224

【参考文献】