基于混沌理论的马尾松毛虫有虫面积BP神经网络预测

BP neural network forecast of Dendrolimus punctata punctata occurrence area based on chaos theory

通过对1983—2010年马尾松毛虫发生数据特点的分析,应用相空间重构技术,将混沌理论和神经网络理论相结合,提出了1种基于混沌神经网络理论的马尾松毛虫有虫面积预测模型。结果表明,该模型有较好的预测能力,当输入层神经元个数(即嵌入维数)为7、隐含层神经元个数为15时,预测未参与建模的2009年越冬代、2010年第1代马尾松毛虫有虫面积的平均相对误差为12.50%。

A model for forecasting Dendrolimus punctata punctata occurrence area was established with the phase space reconstruction technology and combination of chaos theory and neural network theory based on the analysis of the data between 1983 and 2010.The results showed the BP neural network model based on chaos theory has good forecast ability.If there were 7 neurons in input layers and 15 neurons in hidden layers,the average relative error in the forecast of the occurrence areas was 12.50% for the overwintering generration in 2009 and the first generation in 2010,which were not involved in the model construction.