基于支持向量机的房地产上市公司财务危机预警研究

发布时间：2018-01-11 23:02

本文关键词：基于支持向量机的房地产上市公司财务危机预警研究　出处：《西南财经大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着市场经济的发展和金融制度的完善,通过上市实现融资的房地产公司业已达到139家。房地产上市公司经营业绩的好坏与投资者收益的多少是息息相关的。通过研究上市公司财务报表,构建财务危机预警模型对企业存在的财务风险进行评估和预警,有利于企业及时发现风险、控制风险,投资者合理投资、及时止损。财务危机预警理论起源于西方,它通过利用企业财务数据对企业风险进行识别,发现其潜在的财务风险提前进行预警和控制。财务危机预警的方法大致可分为定性分析法和定量分析法两种。定性分析法主要包括个案分析法、标准化调查法等,定量分析法主要包括单变量模型、多变量模型等。随着统计学理论和机器学习的发展,支持向量机作为一个新的方法被应用到财务危机预警领域中来。支持向量机本质上是一个有约束的二次优化问题,作为监督学习的一种,其在分类和预测方面有着广泛的应用。由于它本身所具有的独特优势,支持向量机能够有效地解决识别系统、信用评估等方面的问题。本文借助支持向量机这种机器学习的方法来解决房地产上市公司的财务危机预警问题,为公司财务危机预警提供了新思路、新方法。本文建立在财务危机预警理论和支持向量机理论的基础上,构建基于支持向量机的财务危机预警模型,对房地产上市公司潜在的财务危机进行预测。本文首先介绍了研究背景、研究问题、研究意义与思路,接着对财务危机的基本定义和成因,财务危机预警的定义、意义和方法进行阐述,介绍了国内外学界在财务危机预警领域的研究成果。然后着重介绍了机器学习和支持向量机的基本概念和理论基础,对线性支持向量机和非线性支持向量机各自的算法进行了阐述。接下来建立了财务预警指标体系,对数据进行标准化处理后进行因子分析,提取主因子后构建基于不同参数值的非线性支持向量机,并且对比了不同参数值下的模型预测结果。为了更好地体现支持向量机在财务危机预警精度方面的优势,本文还采用了判别分析和Logistic回归模型分别对样本数据进行财务危机预警,将其得到的预测结果与支持向量机的预测结果进行比较发现,支持向量机的预测结果要明显优于判别分析和Logistic回归的预测结果。借助本文构建的财务危机预警模型,我们可以利用房地产上市公司往年的财务数据对其未来发生财务危机的概率进行预测。对于发生财务危机概率高的企业投资者应该重点关注,谨慎投资；这些企业的管理者应该对企业存在的问题及时进行梳理解决,做到有效控制其财务风险。
[Abstract]:With the development of market economy and financial system, through the implementation of the listed financing Real Estate Company has reached 139. How many quality and investors operating performance of real estate listed companies are closely related. Through the study of the financial statements of listed companies, construct the financial crisis early-warning model for assessment and early warning of enterprise financial risk exists in favor of enterprise timely detection of risk, control risk, investors rational investment, timely stop.
The origin of financial crisis early-warning theory in the west, which based on the enterprise risk identification using the financial data of the enterprise, find the potential financial risk early warning and control method. The financial crisis early warning can be divided into qualitative analysis method and quantitative analysis method. Two kinds of qualitative analysis methods mainly include case analysis, standardized survey method, quantitative analysis method mainly includes single variable model, multi variable model. With the development of the theory of statistics and machine learning, support vector machine is used as a new method has been applied to the field of financial crisis early warning.
Support vector machine is essentially a constrained optimization problem two times, as a supervised learning, which is widely used in classification and prediction. Because its itself has the unique superiority, the support vector machine can effectively solve the recognition system, credit evaluation and other aspects of the problem. In this paper, with the support of this method of vector machine learning to solve the financial crisis early warning of listed real estate companies, provides new ideas and new methods for early warning of financial crisis.
This paper is based on the financial crisis early warning theory and the support vector machine theory, construct the financial crisis early-warning model based on support vector machine, the forecast of real estate listed companies' potential financial crisis. This paper firstly introduces the research background, research questions, research significance and ideas, then the basic definition and the causes of the financial crisis and the definition of the financial crisis early warning, significance and methods are presented in this paper, introduces the research results of domestic and foreign scholars in the field of financial crisis. Then it focuses on the machine learning support vector machine and the basic concepts and theoretical basis of linear support vector machine and support vector machine algorithm of nonlinear respectively are discussed. Next the establishment of the financial early-warning index system, factor analysis on standardized data, extract the main factor after the construction of non linear parameter values based on different support Vector machines, and compare the model prediction results under different parameter values.
In order to better reflect the advantages of support vector machine in the financial crisis early warning accuracy, this paper also uses discriminant analysis and Logistic regression model were used to the financial crisis early warning of the sample data, the prediction results obtained by the support vector machine and comparing the results, the prediction results of support vector machine to predict the outcome of superior discrimination analysis and Logistic regression.
With the help of the financial crisis early-warning model constructed in this paper, we can utilize the financial data of listed real estate companies in previous years to predict the probability of future financial crisis. The financial crisis occurred with high probability of enterprise investors should focus on, prudent investment; the management of these companies should be on business problems to sort out and resolve in a timely manner, do the effective control of the financial risk.

【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F299.233.42

【参考文献】