基于深度学习模型及组合模型的沙漠面积预测研究

Research on desert area prediction based on deep learning model and combination model

沙漠化是全球性的环境问题,影响着许多国家和地区。准确地预测沙漠面积对于制定有效的沙漠化防治策略至关重要。文中使用不同的模型对沙漠面积进行预测,试图找到一种预测准确度高且性能优良的模型。以新疆若羌县东部地区沙漠为研究对象,分别采用ARIMA、RNN、LSTM、GRU、ARIMA-RNN、ARIMA-LSTM、ARIMA-GRU模型对沙漠面积进行预测,使用均方误差、均方根误差和平均绝对误差评估模型的性能。实验发现:ARIMA模型预测准确度最低且模型性能最差,深度学习模型的预测准确度最高可达约96.74%;组合模型的预测准确度最低可达约93.08%,其中ARIMA-GRU组合模型的预测准确度最高约为97.46%。实验表明,深度学习模型在沙漠面积预测中预测准确度高且性能良好,组合模型能够提高沙漠面积预测的准确性和稳定性,能避免单一模型预测的局限性和风险性。

Desertification is a global environmental issue that affects many countries and regions.Accurate prediction of desert area is essential to develop effective desertification control strategies.In this paper,different models are used to predict the desert area to find out a model with high prediction accuracy and excellent model performance.The desert in the eastern part of Ruoqiang County,Xinjiang is taken as the research object and the models of ARIMA(autoregressive integrated moving average),RNN(recurrent neural network),LSTM(long short-term memory),GRU(gated recurrent unit),ARIMA-RNN,ARIMA-LSTM and ARIMA-GRU are used to predict the desert area.The model performance is assessed with mean square error(MSE),root-mean-square error(RMSE)and mean absolute error(MAE).The experimental results show that the model ARIMA is found to have the lowest prediction accuracy and the worst performance,and the deep learning model has the highest prediction accuracy of about 96.74%.The prediction accuracy of the combined model can be as low as about 93.08%,among which the highest prediction accuracy of the combined model ARIMA-GRU reaches about 97.46%.The experiments show that the deep learning model has high prediction accuracy and good performance in desert area prediction.The combined model can improve the accuracy and stability of desert area prediction,and can avoid the limitations and risks of single model prediction.