神经网络在经济预测中的应用

Application of Artificial Neural Network in Economic Forecasts

阐述了针对BP神经网络在经济预测应用中存在的网络学习速度缓慢、计算量大、网络容易陷入局部极小、学习步长需要通过实验由人工来设置和调整等问题,提出将自组织理论、扩展Kalman滤波、模糊算法、小波理论等与神经网络相结合,构成新的网络结构或改进学习算法,以克服BP神经网络自身不足的思路.

This paper analyzes the problems in economic forecasting with BP neural network. The main problem is that the slow speed of network learning, the great capacity of calculation, easily sinking into partial smallest worth,the length of learn step needed to artificially set and adjust through experiment, etc. In order to overcome the shortcomings of BP neural network, we combine neural network with the theory of self-organization, extended Kalman filter, genetic algorithm, wavelet theory etc. and compose new neural network architectures or improve learning algorithm.