Research On Neural Network Models Selection Of Financial Crisis Pre-warning For Listed Companies

With the rapid development of economic globalization and market economy in China, the external environment on which enterprises depend for existence and development is becoming more complex. Enterprises have to face various risks in different stages of production and business activities,which may lead to management difficulties, even to financial crisis or business failure. However, enterprises usually fall into business difficulties little by little. If the wrong practices in business operations and the worsening of the financial situation can be found in time,we can take effective measures and enforce management in advance,thus to avoid the financial crisis. Therefore,it is of great value, theoretically and practically, to establish an accurate and effective financial pre-warning model to discover the potential problems in advance,to help take relevant measures to avoid risks for enterprises.On the ground of the researches by the scholars at home and abroad,firstly, this thesis talks about the definition, characteristics and causes of financial crisis, and then explains the theoretical basis of the pre-warning of it. Secondly, it reviews the research findings of financial crisis cases,analyzing and comparing different pre-warning methods, and summarizes the pros and cons of various methods. Besides the author conducts the theoretical research on introducing the neural network to the pre-warning system, emphasizing the feasibility and necessity of introducing artificial neural network. Through theoretical analysis, the thesis first introduces different neural networks to the pre-warning models and, in comparison, aims to find a more applicable neural network model and to improve the predictive accuracy and operation, thus serving the listed companies effectively.The listed companies in China are the vital forces for other industries, which have great influence on the national economy and matter the interests of the investors as well. The proportion of the manufacturing listed companies in the Shanghai and Shenzhen Stock Markets is very large,which make the pillar industry in the national economy. Therefore, the present thesis, with its focus on the listed manufacturing companies in the two stock markets and reference to the relevant research, chooses 23 financial ratio analyses to study the three-year statistics before the happening of crisis. With the help of SPSS, the financial indicators are carefully selected, thus 14 indicators with significant means are chosen to establish the pre-warning system. Then, Matlab7.0 is employed to set up three neural network pre-warning models of financial crisis and then the author makes comparison among them. The results show that LVQ neural network model the advantage over the BP model in sound judgment and easy operation and some models do not apply to the pre-warning study of financial crisis in China, e.g. PNN neural networks, etc.