摘要
时间序列预测分析方法是进行预测预报的有效工具,有着广泛的应用。针对时间序列的非线性、动态变化等特征,基于RBF神经网络对时间序列预测方法进行改进,并以安徽省池州市1959～2009年来的月降水量为时间序列数据样本,用MATLAB软件编程,采用基于随机选取中心的RBF神经网络预测方法,对池州市的月降水量进行预测,并选择不同的扩展速度参数,用均方误差进行检验。通过与BP网络模型的预测结果比较分析,表明RBF模型的预测效果较好。建立的基于随机选取中心的RBF神经网络模型,不需要计算原始时间序列数据的复杂函数关系,具有操作简单、学习速度快、短期预测精度高等优点,用于时间序列预测方面能够获得十分满意的结果,具有很高的应用价值。
The time series analysis which used to forecast time series is an effective method and has been used widely. For the non-linear and dynamic characters, we improved time series forecast method based on RBF neural network with month precipitation of 1959-2009 in Chizhou, Anhui as data sample of time series. We used MATLAB software to program and forecast month precipitation of Chizhou based on RBF neural network fore- cast method of random selection center. At last, we chose different speed and test it with RMSE （root mean square error） . Compared with BPNN（backpropagation neural network）, predictive validity of RBF model based on ran- dom selection center is preferable and does not need to calculate complexed functional relationship of the original time series data, with advantage of simple operation, study fast and high short-term forecast accuracy.
出处
《安徽农业大学学报》
CAS
CSCD
北大核心
2012年第3期451-455,共5页
Journal of Anhui Agricultural University
基金
国家自然科学基金项目(70271062
40771117)
安徽省高校省级重点科研基金项目(KJ2010A121)共同资助
