基于自适应神经模糊推理系统的汽车制造业财务预警研究

发布时间：2018-04-23 12:08

本文选题：汽车制造业 + 财务预警　；参考：《广西科技大学》2015年硕士论文

【摘要】：制造业是国民财富的源泉,而汽车制造业又是其中的重中之重。尽管我国汽车制造业企业的规模和数量都在日益壮大,但是整个行业抵抗风险的能力却相当薄弱,极易受到总体经济环境和国家相关政策的影响,进而引发严重的财务危机。因此,构建汽车制造业财务预警模型,提升汽车制造业应对财务危机的能力具有十分重要的意义。本文首先介绍了研究的背景、理论及现实意义,在借鉴国内外学者在财务预警方面的研究成果的基础之上,提出了本文的研究方法、思路及创新之处。然后介绍了本文所认定的财务业预警的相关概念,ANFIS及本文中采用的3种聚类算法以及ANFIS和聚类算法的结合方法。接着,在参考前人研究假设的基础之上,明确了汽车制造业财务预警的影响因素,包括财务方面和非财务方面。财务因素有现金流量、长短期偿债能力、运营能力、成长能力、盈利能力七个因素,非财务因素有股权结构、公司治理两个因素。财务预警指标体系是预警模型构建的基础,也是直接影响预警效果的重要因素之一。为了更全面、客观、真实反映汽车制造业的财务状况,偿债能力、现金流量能力、盈利能力、成长能力及运营能力等方面的26个财务指标用来衡量财务因素,第一大股东持股比例等7个非财务指标来构建财务预警指标体系。在实证研究部分,本文从深圳和上海证券交易所中A股中选取了54家汽车整车制造或零部件制造上市企业以构建研究所需的样本集,其时间跨度为11年(2003-2013)。在该样本集中,三分之一由财务危机企业构成,三分之二是与之配对的非财务危机企业。对所有样本的上述33个财务和非财务指标同时进行显著性检验和主成份分析,剔除了某些相关性不明显的财务和非财务指标,并对所筛选的财务指标进行了降维处理,最终得到5个综合财务指标因子和4个非财务指标,从而构建了我国汽车制造业上市公司财务预警指标体系。利用上述指标体系及样本集中的数据,运用基于聚类算法的ANFIS方法构建汽车制造企业的预警模型和BP神经网络、SVM(支持向量机)及Logit回归这三种常用的预警模型进行建模,通过模型的检验、比较分析可以得出以下结论:基于聚类算法的ANFIS的汽车制造业上市公司财务预警模型明显优于其他模型,此外该模型与BP神经网络模型及SVM(支持向量)等人工智能模型一道,明显优于Logit回归这一非人工智能模型。通过上述研究,本文针对汽车制造业上市公司财务预警实践提出了两点建议:(1)构建汽车制造业财务预警指标体系时,应剔除某些干扰指标。(2)选取建模工具时,应优先考虑人工智能建模预测工具。
[Abstract]:Manufacturing industry is the source of national wealth, and automobile manufacturing industry is one of the most important. Although the scale and quantity of automobile manufacturing enterprises in our country are growing day by day, the ability of the whole industry to resist the risk is quite weak, which is easily affected by the overall economic environment and the relevant policies of the country, and then causes the serious financial crisis. Therefore, it is of great significance to construct the financial warning model of automobile manufacturing industry and improve the ability of automobile manufacturing industry to deal with financial crisis. Firstly, this paper introduces the background, theory and practical significance of the research, and puts forward the research methods, ideas and innovations of this paper based on the research results of domestic and foreign scholars in financial early warning. Then this paper introduces the related concepts of financial industry early warning, such as ANFIS, three clustering algorithms used in this paper, and the combination of ANFIS and clustering algorithm. Then, on the basis of the previous research hypothesis, the paper clarifies the influencing factors of automobile manufacturing financial early warning, including financial and non-financial aspects. Financial factors include cash flow, long-term and short-term solvency, operation ability, growth capacity, profitability, non-financial factors have equity structure, corporate governance two factors. The financial early warning index system is the foundation of the early warning model and one of the important factors directly affecting the early warning effect. In order to more comprehensively, objectively and realistically reflect the financial situation, solvency, cash flow capacity, profitability, growth ability and operation ability of the automobile manufacturing industry, 26 financial indicators are used to measure financial factors. The first major shareholder holding ratio and other seven non-financial indicators to build a financial warning index system. In the part of empirical research, this paper selects 54 listed enterprises of automobile manufacturing or parts manufacturing from the A shares of Shenzhen and Shanghai Stock Exchange to construct the sample set needed for the research, whose time span is 11 years from 2003 to 2013. In this sample set, 1/3 is composed of financial crisis enterprises and 2/3 is a matched non-financial crisis enterprise. The above 33 financial and non-financial indicators of all samples were tested simultaneously with the principal component analysis, and some uncorrelated financial and non-financial indicators were excluded, and the selected financial indicators were treated with dimensionality reduction. Finally, five comprehensive financial index factors and four non-financial indicators are obtained, and the financial early-warning index system of Chinese automobile manufacturing listed companies is constructed. Using the above index system and the data of sample set, using the ANFIS method based on clustering algorithm to build the early warning model of automobile manufacturing enterprise and BP neural network SVM (support vector machine) and Logit regression three commonly used early warning models to model. Through the test of the model, the comparative analysis can draw the following conclusions: the financial early-warning model of automobile manufacturing listed companies based on clustering algorithm ANFIS is obviously superior to other models. In addition, this model, together with BP neural network model and SVM (support vector) artificial intelligence model, is obviously superior to the non-artificial intelligence model such as Logit regression. According to the above research, this paper puts forward two suggestions on the financial early warning practice of listed companies in automobile manufacturing industry: 1) when constructing the financial warning index system of automobile manufacturing industry, we should eliminate some interference indexes. 2) when we select the modeling tools, Priority should be given to artificial intelligence modeling and prediction tools.
【学位授予单位】：广西科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F426.471;F406.7

【相似文献】