作　　者： ;

机构地区： 广东培正学院经济学系

出　　处： 《广东培正学院学报》 2013年第4期7-11,共5页

摘　　要： 采用动量算法对BP神经网络进行优化并对广州市GDP增长率进行预测，取得了理想的效果。首先建立了标准的基于梯度下降算法的BP网络和基于动量算法优化的BP网络，然后建立了传统的移动平均模型及指数平滑模型，最后对四个模型的误差结果进行了对比。研究表明：动量优化网络的预测精度显著优于其他三种方法，是一种值得推荐的预测GDP增长率的方法。 Based on momentum algorithm, this paper constructs an optimized BP neural network to forecast the GDP growth of Guangzhou and achieves satisfactory results. At first, a BP network is built on the ground of standard gradient descent algorithm, and an optimized BP network is done on the basis of momentum algorithm. Then, a traditional moving average model and an exponential smoothing model are also established. At last, er- rors of the four models are contrasted. The study shows that the momentum algorithm BP network is significantly better than the other three models in prediction accuracy and is thus recommended.

关 键 词： 神经网络 标准梯度下降 动量 预测

领　　域： [经济管理]