摘要
BP神经网络具有较强的非线性问题处理能力,是目前一种较好的用于时间序列预测的方法,然而它存在易于陷入局部极小值、收敛速度慢等不足。针对以上缺点,利用改进粒子群优化BP神经网络的权值与阈值,有效地增强了算法的全局搜索能力和提高了收敛速度。针对地震预测的应用,用改进粒子群优化的BP算法对四川地区最大震级时间序列进行预测,通过训练预测次年的最大震级。结果表明此方法优于未经优化的算法,具有良好的预测效果。
Back propagation neural network is a good method used in time series prediction, which has a strong nonlinear capacity of processing problem. However, it is easy to fall into local minimum and slow convergence. To solve the above shortcomings, using improved particle swarm algorithm optimize the weights and threshold of back propagation neural network can effectively enhance the ability of the algorithm＇s global search and improve the convergence rate. For applications of earthquake prediction, using an improved particle swarm of the BP neural networks optimization algorithm to predict the time series of largest magnitude in Sichuan Province. Predict the largest seismic magnitude of the following year by training. Results show that this method is better than the BP algorithm without optimization and has good result of prediction.
出处
《计算机与数字工程》
2013年第2期155-157,290,共4页
Computer & Digital Engineering
基金
高分辨率对地观测重大科技专项项目(编号:E0311/1112/JC02)资助
关键词
BP神经网络
粒子群优化算法
时间序列
地震预测
BP neural network, particle swarm optimization algorithm, time series, earthquake prediction
