Combination Classifier Was Applied Research, Financial Early Warning

Financial warning system to prevent the financial system is running deviation anticipated goal and the establishment of the alarm system, Financial warning system have specific aim and predictive characteristics. It through the indexes and comprehensive assessment to predict the financial status, development trends and changes, to provide the scientific decision the strength of support for policy makers.The development of artificial neural network for several years. In addition to the neural network theory has been further development and the application of the neural network are also increasingly rich. Neural network for the future development of a direction is to successfully applied in all fields, play the information processing power, expand their application scope.This paper takes listed companies as the research objection, financial warning theory and model was studied and discussed. This paper first to subject research at home and abroad in the field of literature, sorting and summarized, to the financial warning the development of related technologies, makes a brief analysis. Then expounds the principle and method of classifier, and introduced the classifier set the two most popular Adaboost algorithm and Bagging, analyzes their advantages and disadvantages. Choose the BP neural network classifier, as a combined the classifier, and neural network integration puts forward the main purpose is to solve a single network existing problems, and give a specific principles and algorithms, to the financial warning process of the system in this paper.According to the selected learning algorithm to select1350listed companies of the financial data for the corresponding processing, to the financial data of ten can compare to reflect the company's financial situation indexes, and study the BP single classifier different parameters in the performance difference and Adaboost strong learning algorithm combined the capability of classifier under differences, experimental results show: combination forecast model and the traditional classification compared single classification model, with strong superiority, can effectively improve the accuracy.The results of simulation experiment to reduce the risk of a financial accident company, and also provides some reference methods and measures.