Web信息驱动的上市公司财务危机预警研究

发布时间：2018-01-14 19:14

本文关键词：Web信息驱动的上市公司财务危机预警研究　出处：《江西财经大学》2013年博士论文　论文类型：学位论文

【摘要】：市场经济的发展,使得公司间的竞争日益激烈,而全球经济一体化,带来的不仅仅是发展机遇,也暗藏了无尽的危机和风险。对上市公司而言,因为发生财务危机而被特别处理甚至被迫退市,不仅影响其自身的生存和发展,还会给投资者和债权人带来巨大的经济损失。因此,对上市公司财务危机准确、及时、有效地预警,无疑对促进资本市场和国民经济的发展,维持社会稳定具有重要作用。财务危机预警研究主要涉及两个方面：预警指标和预警模型。已有的研究在预警指标方面主要是选用财务指标,然而由于财务指标的滞后性和易被操纵等固有缺陷,也有学者积极引入如宏观经济变量、公司治理变量、产业变量等非财务指标,但是由于非财务指标类型众多,数据不易获取,而且有些指标难以量化,这些都为在财务危机预警中非财务指标的引入造成障碍。而由于网络技术的发展,Web金融信息大量涌现,其所具有的实时性、覆盖性、全面性和易获取等特点,正好弥补了非财务指标获取困难以及不完整、不全面等不足,为财务危机预警中非财务指标的获取提供了新的途径。对财务危机预警模型的研究,已有的研究成果从传统的统计模型到人工智能模型,大多是用截面数据所建立的静态预警模型,而上市公司的财务危机并不是突然发生,有一个逐渐演化的过程,静态预警模型没有考虑到预警指标的时序特征,忽略了历史数据对预警结果的影响,从而造成预警模型的早期预警效果较差,在实际应用中难以推广。本文从以上两个方面入手,一方面,研究如何将Web金融信息引入上市公司财务危机预警指标体系以及Web金融信息指标的预警作用；另一方面,研究上市公司财务危机的动态预警。围绕这两大方面,本文具体研究了以下内容： (1)研究了Web金融信息的量化问题。因为Web金融信息基本上是非结构化的文本信息,所以,要将它纳入到财务危机预警指标体系,需要将它进行合理地量化,文本内容的情感倾向值计算是文本信息量化的常用手段。本文针对Web金融信息文本,构建了金融领域情感词典,提出了基于语素分数的情感倾向值计算方法。 (2)分析了Web金融信息和上市公司财务状况的关系。针对Web金融信息和上市公司财务状况的关系,本文主要研究了两个方面。首先,通过相关性分析研究了Web金融信息指标(即信息热度和情感值)与财务指标的关系；其次,运用Logistic回归分析研究了Web金融信息指标对预警上市公司是否会被ST的影响。 (3)验证了Web金融信息指标对财务危机预警模型的影响。本文运用LIBSVM分别构建了纯财务指标预警模型和财务指标与Web金融信息指标相结合的混合模型,通过实证比较分析,发现加入Web金融信息指标后,预警模型在超前性、稳定性和有效性等方面都有很大程度的改善。 (4)研究了上市公司财务危机的动态预警。本文首先运用计量经济学的ARMA模型,对上市公司财务状况的时序特征进行拟合；然后借用质量管理学上的控制图思想,将Web金融信息情感褒贬程度指标加入EWMA,构建了上市公司财务危机动态预警模型S-EWMA;最后对其进行实证分析,与EWMA进行了对比分析。本文的创新性工作体现在： (1)提出了基于语素分数的Web金融信息文本情感倾向值计算方法。目前,文本情感倾向性计算方面已有不少的研究成果,但是计算方法往往受到种子词选择和情感词典覆盖性等方面的限制,而且专门关于金融领域文本情感计算的研究尚未发现。本文所提出的基于语素分数的Web金融信息文本情感倾向值计算方法,具有领域针对性,能充分满足金融领域文本情感倾向性分析的要求。首先,构建了金融领域的情感词典,将金融领域的特色情感词添加到词典中,而基于语素分数的情感值计算方法很好地解决了情感词典的覆盖性问题；其次,在进行句子和文档的情感倾向值计算时,充分考虑文档结构中的否定词和程度副词对文档情感倾向所起的修饰作用,考虑了句子在文档不同位置对情感倾向的贡献不同,从而被赋予不同权重,以及子句间连接词的转折、递进、并列等模式对句子情感倾向的影响,而不是将各组成部分的情感值进行简单地求和。实验结果也验证了本文计算方法的有效性。 (2)分析了Web金融信息和上市公司财务状况的相关性。本文通过对Web金融信息和上市公司财务状况的关系分析,发现Web金融信息文本情感值中含有财务指标未曾包含的与上市公司财务状况相关的信息,因此Web金融信息情感值可以作为上市公司财务指标的重要补充。 (3)构建了财务指标与Web金融信息指标相结合的上市公司财务危机预警模型。本文结合Web金融信息指标和财务指标构建了混合指标的上市公司财务危机预警模型,实证结果表明,该模型在预警的有效性、稳定性和超前性等方面均优于纯财务指标模型。 (4)构建了加入Web金融信息情感褒贬程度指标的上市公司财务危机动态预警模型S-EWMA.该预警模型基于财务指标的动态面板数据ARMA模型而构建,能很好地反映财务指标的时序特性,又在指数加权移动平均控制图中加入了Web金融信息情感褒贬程度指标,弥补了财务指标滞后性等缺陷,能够反映上市公司财务危机逐步演变发展的动态性及演变趋势,能够有效地预警财务危机发生的时点。实证分析表明,该模型可以较大地提高财务危机预警的超前性。
[Abstract]:The development of market economy, the competition between companies is increasingly fierce, and the global economic integration, not only bring development opportunities, but also hidden endless crises and risks. For listed companies, and even forced to withdraw from the market because of the special treatment by the financial crisis, not only affect their own survival and development, but also bring huge the economic losses to investors and creditors. Therefore, the financial crisis of the listed companies is accurate, timely, effective early warning, is to promote the development of capital market and the national economy, plays an important role in maintaining social stability.
Study on early warning of financial crisis mainly involves two aspects: the early warning indicators and warning model. The existing research on the early warning index is the choice of the main financial indicators of financial indicators, however due to the lag and easy manipulation of inherent defects, some scholars actively introduce such as macroeconomic variables, corporate governance variables, non-financial indicators of industry variables however, because of the many types of non-financial indicators, data acquisition, and some indicators are difficult to quantify, these are in the early warning of financial crisis in the non obstacle. The introduction of financial indicators and the development of network technology, the emergence of a large number of Web financial information, real-time, it has the characteristics of comprehensive coverage, and easy get, just to make up the non financial indicators are difficult to obtain and not complete, is not comprehensive, and provides a new way for the acquisition of non-financial indicators of financial crisis early warning.
Study on financial crisis early warning model, the existing research results from the traditional statistical model to artificial intelligence model, mostly static prediction model built using cross section data, and the listed company's financial crisis is not a sudden, there is a gradual evolutionary process, static early-warning model does not consider the temporal characteristics of early warning indicators, ignoring the influence of historical data of the early warning results, poor early warning result of the early warning model, it is difficult to promote in the practical application.
This article from the above two aspects, on the one hand, to study how the early warning function Web financial information into the listed company's financial crisis early warning index system and Web financial information index; on the other hand, the dynamic early warning of financial crisis of listed companies. Based on these two aspects, this paper studies the following contents:
(1) the quantification study on Web financial information. Because the Web financial information is basically unstructured text information, so, to put it into the financial crisis early warning index system, it needs to be reasonably quantified, text sentiment content value calculation is a common method of text information based on quantization. Text Web financial information, constructs the financial sector sentiment dictionary presents value calculation method based on fractional morpheme sentiment.
(2) analyzes the relationship between the financial condition of Web financial information of listed companies and the relationship between the financial condition of Web. And financial information of listed companies, this paper mainly studies two aspects. First, through the correlation analysis of the Web financial information index (i.e. information heat and emotional value) relationship with financial index; secondly, the use of Logistic analysis of the Web financial information index on early warning of listed companies will be the impact of ST regression.
(3) examined the effect of Web financial information index of the financial crisis early warning model. LIBSVM were constructed by the hybrid model of pure financial early-warning model and financial index and Web index of the combination of financial information used in this paper, through empirical analysis, we found that the addition of Web financial information indicators, early warning model in advance, stability and efficiency other aspects are greatly improved.
(4) the dynamic early warning of financial crisis of listed companies. This paper uses ARMA model of econometrics, the timing characteristics of the listed company's financial position was fitting; then using the control chart thought on quality management, Web financial information emotion appraise index of the degree of joining EWMA, constructed the S-EWMA dynamic financial crisis warning model of listed company; finally carries on the empirical analysis, comparing with EWMA.
The innovative work of this article is reflected in the following:
(1) put forward the calculation method of morpheme scores of the Web financial information based on the text sentiment. At present, research results of sentiment computation are many, but the calculation method is often affected by seed word selection and coverage of sentiment dictionary and other aspects, and specialized research on emotional financial field text computing has not been found. The proposed method of calculating the morpheme scores of the Web financial information based on the text sentiment, with the field of pertinence, can fully meet the analysis of financial domain sentiment requirements. First, build the financial sector sentiment dictionary will feature the emotional words in the financial field is added to the dictionary, and the scores of emotional morpheme based on the value calculation method solves the problem of covering the emotion dictionary; secondly, in the emotional tendency of sentence and document value calculation, considering the Modified the role of negative words and adverbs file structure of the document sentiment, consider the sentence in the document in different positions of different contributions to the emotional tendencies, which are given different weights, and the connection between the word clause turning, progressive, parallel mode effects tend to love the sense of the sentence will not. Simply calculated and the value of each part of the emotion. Experimental results also verify the validity of this method.
(2) to analyze the correlation between financial information and financial status of Web listed companies. Through the analysis on the relation between financial information and financial status of Web listed companies, Web financial information contained in the text emotion value and financial indicators did not contain information related to the financial situation of listed companies, so the Web financial information emotion value can be used as an important supplement financial indicators of listed companies.
(3) construction of the listed company's financial crisis early warning model of financial indicators and financial indicators combined with the Web information. To construct the financial crisis warning model of listed companies based on the hybrid index Web of financial information and financial index, the empirical results show that the model is effective in early warning, stability and advance is superior to pure the financial index model.
(4) the construction of joining the Web financial information likeness degree of listed company financial crisis early warning model of the S-EWMA. dynamic early-warning model and build dynamic panel data ARMA model based on financial indicators, can well reflect the timing characteristics of financial indicators, and the exponentially weighted moving average control chart with Web financial information likeness the degree of index, make up the lag of financial indicators and other defects, and can reflect the dynamic evolution trend of listed company financial crisis gradually evolved, can effective early warning of financial crisis point. The empirical analysis shows that the model can advance the financial crisis early warning is greatly improved.

【学位授予单位】：江西财经大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：F832.51;F275

【参考文献】