应用人工神经网络预测马尾松毛虫的有虫面积

运用人工神经网络的原理和方法，选取与马尾松毛虫发生量相关关系密切的8个气象因子作为样本的输入特征，建立马尾松毛虫有虫面积与气象因子的BP网络模型，结果表明：所建立的BP模型，具有令人满意的拟合精度和预测精度。当隐含层神经元个数为15个，输入因子数为8个时，18组有虫面积的平均拟合精度为100％，相关系数为1．0000，2组预留有虫面积的平均预测精度为96．85％，预测准确率为100％。

Forecasting Insect Appearance Areas of Dendrolimus Punctatus Walker by Applying Artificial Neural Network (ANN)

We chose eight meteorological factors, which highly related to the occurrence of dendrolimus punctatus walker, as the input characteristics of samples, by applying the theory and methods of the artificial neural network (ANN), a Back Propagation (BP) forecast model between the occurrence of dendrolimus punctatus walker and the meteorological factors was established. The results showed that the established BP model had both satisfying fitting and forecasting precision. With 15 hidden layer neurons and 8 input factors, the average fitting precision for 18 groups of insect appearance areas was 100%, and the coefficient was 1. 0000. The average forecast precision for 2 prepared groups of insect appearance areas was 96.85% , and forecast accuracy was 100% .