Adaptive Optimization of Regional Soil Moisture Prediction Model via Transfer Learning

This study focuses on the adaptive optimization problem of regional soil moisture prediction model and proposes an optimization algorithm based on transfer learning. This study uses transfer learning to extract effective features from rich data in the source domain and transfer them to the soil moisture prediction task in the target domain. The feature weights are adaptively assigned according to the similarity between the source domain and the target domain to improve the adaptability of the model to the target area. The experiment selects multiple different geographical regions as target domains and the regions with a large amount of data as source domains. The root means square error (RMSE), mean absolute error (MAE) and determination coefficient (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) are used as evaluation indicators. The results show that compared with the traditional prediction model, the RMSE of the optimized model in this study is reduced by an average of 15.6%, MAE is reduced by 12.8%, and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> is increased by 0.11. This shows that the new model has higher accuracy and better adaptability in soil moisture prediction in different regions, and provides an effective method for accurate regional soil moisture prediction.