摘要
针对机场短时局地天气预报这一特定的时序数据预测问题,提出用遗传算法同时优化神经网络的连接权值、阈值和网络结构,建立了基于遗传神经网络的天气预报模型,并利用某气象站的实际观测数据进行了实验。仿真结果表明,相对于单独使用BP神经网络,运用该方法可以有效提高预测精度。
During actual application of BP neural network,there exist some problems,including difficult determination of the size and structure of neural network in advance,and it＇s easy to converge to local optima.This paper proposed to use genetic algorithms to optimize the weights and thresholds and the topology structure of neural networks simultaneously,constructed a weather forcast model based on genetic neural networks,and experimented it with some weather data observed by a weather station.The result of the simulation indicated that this method can improve the accuracy,comparing to the BP neural networks.
出处
《软件导刊》
2012年第1期15-17,共3页
Software Guide
关键词
遗传算法
神经网络
天气预报
时序数据预测
Genetic Algorithms
Neural Network
Weather Forcast
Time-Series Data Prediction
