基于神经网络集成的作物需水量预测

Prediction of Crop Water Requirements Based on Neural Network Ensemble

应用神经网络集成模型,以空气湿度、温度、太阳辐射以及风速为输入,利用交叉验证方法确定网络隐层节点数,建立作物需水量的预测模型。实验结果表明,与单个神经网络与随机森林模型相比,神经网络集成模型能获得更好的预测精度,可用于节水灌溉。

In order to predict crop water requirements, a method based on neural network ensemble is developed in this paper. Solar radiation, air temperature, relative humidity and wind speed are used as input variables. The number of hidden-layer neurons is determined by cross validation method. Experimental results indicate that the method based on neural network ensemble outperforms those based on a single neural network and random forest, and can be used for water-saving irrigation.