Improved Bp Neural Network Applications, Commodity Exports Are Forecast

Artificial neural network(ANN) is a nonlinear system which simulates biological neural network using computer. It has powerful self-organizing, adaptive learning, parallel processing and high degree of fault-tolerant capability. So far, many scholars have proposed dozens of neural network models and have applied them to many fields successfully. As the most widely used neural network, BP network has the advantages of simple structure and mature technology. However, BP algorithm is based on gradient descent method, thus it has two main shortcomings of slow convergence rate and easy to fall into local minimum.In this paper, we firstly make a detailed analysis of development process, basic principle, learning methods, classification and application of ANN, and then discuss principle, learning process and shortcomings of BP algorithm, and introduce the status of BP improved algorithm.On the basis of detailed analysis of BP algorithm, we propose a comprehensive improved BP algorithm which associates adaptive learning rate method with dynamic activation function method. During each of learning process of this algorithm, learning rate and activation function will be adjusted dynamically according to error change. Experimental results show that the proposed comprehensive improved BP algorithm can effectively improve the rate of the algorithm convergence.Finally, we apply improved BP neural network into forecast commodity export. We build commodity export forecast model based on improved BP neural network to forecast commodity export, and make a comparative study with other common forecast methods. The results show that the accuracy of improved BP neural network model is higher than other methods.