Integrated Neural Network And Its Application In Financial Crisis Forecast

Forecasting of financial crisis is one of the most important forecasting systems in enterprise management, which can help the enterprise find financial crisis in advance through analyzing each financial index, then adjust the management strategy in time. In this paper, main existing forecasting models are investigated and it is found that they have three weaknesses. By combining of neural network and integrated learning technology, a new type of integrated neural network is proposed to solve the problems of forecasting of financial crisis. The experiment proves that if the individual neural network's result is better than the random conjecture, the error rate of integrated neural network is less than the individual neural network's.Meanwhile, it is found that the errors' classification is worth our attention. The cost of considering the financial health as the financial crisis is to strictly check and financially supervise the enterprise, and the cost of considering the financial crisis as the financial health is to ignore the worse financial situation, then the enterprise cannot adjust the management strategy until the financial crisis takes place. So the risks of these two wrong judgments are completely different, the former's risk is smaller than the latter's. According to this point, this paper changes from the traditional concept based on the minimal classification error rate, to the integrated neural network based on the minimal risk (cost) to forecast the financial situation. Based on indexes such as FPR, FNR, and initial weight setting and dynamic weight adjusting, the proposed integrated neural network is able to pay more attention to FPR and small samples. Under steady general error rate, experiments prove that reducing the latter's error rate is much meaningful.