基于LSTM递归神经网络的番茄目标产量时间序列预测

Time Series Prediction of Tomato Yield Based on LSTM Recurrent Neural Network

番茄目标产量预测对于合理制定灌溉施肥制度以及减少水肥的浪费有着重要意义。以番茄历年产量数据为依据,提出了一种基于长短期记忆递归神经网络的番茄目标产量预测模型,包括模型设计、网络训练和预测过程实现等,将模型预测结果与自回归移动平均(ARIMA)、小波神经网络(WNN)、支持向量回归(SVR)3种时间序列预测模型进行对比,验证了所提出的LSTM递归神经网络预测模型在番茄目标产量预测中具有较高准确性。

The prediction of tomato target yield is of great significance to reasonably make plans of irrigation and fertilization and to reduce the waste of water and fertilizer. In order to improve the accuracy of yield prediction, a new tomato target yield prediction model based on long short-term memory （LSTM） and recurrent neural networks （RNN） is proposed in this paper, including model design, net training and prediction process realization. The prediction result of the proposed model is compared with that of three time series prediction models, including ARIMA, WNN and SVR. It is verified that the proposed LSTM recurrent neural network has high accuracy in the prediction of tomato target yield.