Research On Financial Diagnosis Of Listed Companies Based On Data Mining

By the time of September 2015, there are 2,883 listed companies in China’s Shanghai and Shenzhen stock exchange, in which there are 1,828 manufacturing listed companies. Thus, the manufacturing listed company plays an important role in our national economy. The healthy development of manufacturing listed companies make an important contribution to boosting employment, increasing government revenue, promoting the GDP growth and the development of relevant industries. Therefore, the importance of avoiding the listed companies sliding into financial crisis and the effective control of listed companies’ financial risks are self-evident.By the empirical research on financial diagnosis, this thesis establishes the index system that is suited to Shanghai and Shenzhen Stock Exchange’s A-share listed companies in the manufacturing industry, and also this thesis provides a reference for other industries. Through the theoretical study of financial crisis, financial diagnosis and listed companies in manufacturing industry, this thesis specifies the roots of manufacturing listed companies’ financial crisis in China, various signs of financial crisis which is going to happen and how to make the appropriate precautions to the internal and external of the listed companies.The main purpose of this thesis is to use the financial data in 2011 to predict the financial situation of 160 sampling manufacturing listed companies in 2014 and to determine whether these companies in 2014 have the risk of delisting warning(* ST). This thesis employs data mining tools flexibly, clustering analysis in statistical analysis tools as well as gray correlation analysis in knowledge discovery tools. From both the relevance and importance, this thesis screens financial indicators for the financial diagnosis and constructs a more precise index of financial diagnostic indicators. Afterwards, in the light of three commonly used financial forecasting methods in financial diagnosis field--Logistic regression, CART decision tree and BP neural networks, the sifted index will be adopted into these three models in turns to predict. The results show that in terms of *ST companies’ prediction, the accuracy rates of Logistic regression model and CART decision tree model are lower than BP neural network model; and the accuracy rates are 70% and 75% respectively, while the accuracy rate of BP neural network reached to 80%. In the scope of healthy companies’ prediction, the accuracy rates of Logistic regression model and CART decision tree model are both higher than the BP neural network model, and the accuracy rates reach to 86.67%, and 88.33% respectively, while the accuracy rate of BP neural network model is only 73.33%. From the above results, after being filtered through clustering and gray relational analysis method, the predicting effects of these three models are pretty good. Besides, these three models have their own strengths in predicting *ST companies and healthy companies. In practical application, researchers should choose one of these models reasonably. And we can also use these methods in combination to improve the prediction accuracy rate.