Research On Enterprise Financial Early Warning Of BP Neural Network Optimized By Modified HS Algorithm

For a long time, the research of enterprise financial warning is the prolonged topic of every field in the capital market. With the existing research results and the development of application requirements, it is not difficult to see that financial crisis early warning follows the empirical research of the enterprise financial, and also need the management process which is complex and comprehensive. It also need warning and process control theory, multi-subject knowledge as dynamic information technology and the like, and it is inseparable from the support of efficient prediction model and prediction technology. An effective financial crisis early warning system can provide comprehensive precise and real-time financial information to prepare to alert. Therefore, it is necessary to further study the financial early warning mechanism, at the same time,we should combine the special case of the capital market in our country to create financial early warning model which is suitable for our real-time situation from the learning process of foreign mature research results. When we insight into the sign of the financial crisis,we can warn signals in advance for operators and investors to plan, also can avoid the further expansion of the financial crisis, and summarize lessons learned. In order to improve the accuracy of the enterprise financial crisis warning effectively, and at the same time aimed at the defects existing in the learning process of the error back propagation neural network, such that it has slow convergence speed and fall into local minimal value easily, etc, puts forward a model based on harmony search combined with differential evolution of BP network optimization algorithm to warn the enterprise financial crisis. The variation mechanism of the differential evolution algorithm is used to improve the harmony search algorithm data processing operations so that the search performance of harmony search could be improved;then use the harmony search improved by differential evolution algorithm to complete training after its weights and thresholds of BP were optimized, and get a HSDM-BP model. Take advantage of the trained model for the same amount through ST special treatment or healthy of financial data for training analysis and early warning of listed companies,and the results will becompared with other algorithms. Experiments show that using the model algorithm of this paper to predict the results of the enterprise financial crisis are better than other model algorithms on the prediction accuracy of enterprise crisis divide. The results show that in this paper HSDM-BP network algorithm proposed not only overcome the shortcomings of BP network, improves the accuracy of the financial early warning, while compared with other early warning method is superior in performance.