Financial Distress Early-warning Of Listed Real Estate Companies:Empiric Study Based-on Support Vector Machine

With the improvement of the market economy and the financial system, the number of listed real estate companies which want to achieve funds by stock listing has reached139. Returns obtained by investors are closely related with the quality of results of operations of listed real estate companies. After researching financial statements of listed companies, assessing and warning financial risk which is existed in the listed real estate companies through building financial crisis model. This will help enterprises to control risk and investors to reduce losses.Based on the theory of financial crisis warning and support vector machine, this paper builds financial distress early-warning model based on support vector machine to predict potential financial crisis of listed real estate companies. At first, this paper introduces the research background, research questions, significance and ideas. Then, it discusses the basic definition of financial crisis and the causes, definitions of financial crisis early warning and its meaning and methods. Next, this paper shows the basic concepts and theoretical basis of machine learning and support vector machines, discusses the differences of linear and non-linear support vector machine. After establishing financial early warning indicator system and processing data, this paper uses factor analysis to extract the main factor. At last it builds nonlinear support vector machines based on different parameters and compares the predictions of the model under different parameter values.In order to better reflect the advantage of SVM in the prediction accuracy of financial crisis warning, the paper also uses discriminant analysis and Logistic regression models to predict financial crisis with the same sample data, through comparing the predictions of three models, this paper find that SVM is significantly better than discriminant analysis and Logistic regression.