修正的BP神经网络森林火灾成灾面积预测研究

Study on Prediction of Forest Fire Disaster Area with Modified BP Neural Network

以1997—2012年《中国林业统计年鉴》的全国森林火灾成灾面积为依据,应用BP神经网络模型对成灾面积进行了预测,对预测方法进行了检验。在此基础上,利用残差提出了修正的BP神经网络模型,并对预测方法进行了改进。研究结果表明,修正的BP神经网络预测精度高于BP神经网络,预测相对误差平均为7.2%,可应用于森林火灾成灾面积的预测。

Based on the data of national forest fire disaster area from China Forestry Statistical Yearbook from 1997 to 2012, applyed BP neural network model to predict the disaster area and the prediction method was tested. On this basis, an improved prediction method named modified BP neural network was proposed by using residual error. The results showed that the prediction accuracy of modified BP neural network was higher than BP neurtal network. The modified BP neural network which mean prediction relative error was 7.2 % could be used to predict the forest fire disaster area.