改进的BP算法在地震灾害损失预测中的应用

神经网络因其自组织性、自学习性与高度非线性广泛应用于各种预测领域。针对传统神经网络存在的收敛速度慢,容易陷入局部最小等缺陷,为提高神经网络传统BP算法的训练速度,以3层神经网络为例,通过对权值的分析与优化,推导出改良的BP算法———双权值迭代优化法,将该算法应用于地震灾害损失的预测中去并取得较好效果。

The application of earthquake disaster loss estimation based on improving BP algorithm

The neural network is widely applied in each kind of forecast domain because of its organization sense, studies independently and highly misalignment. In view of the flaws such as lower convergence rate of tradition neural network, easy to fall into partially smallest and so on, to enhance the training speed of the neural tradition BP algorithm network, take three layer neural networks as the example, through the analysis and the optimization for weight value, infers the improved BP algorithm, that is, double weight iteration optimization method. Applies it into the forecast of the earthquake disaster loss and makes better result.