Research and application of rainfall prediction method based on PLS-LSTM modeling

Rainfall prediction has important application value in the fields of agriculture, water resource management, disaster prevention and control, and climate modeling, but traditional methods have limitations in capturing the spatio-temporal complexity of rainfall. Deep learning techniques have become an important tool for rainfall prediction due to their powerful modeling capabilities. In this study, a PLS-LSTM model combining PLS and LSTM is proposed to extract the key features through PLS dimensionality reduction and model the time series dependence using LSTM to significantly improve the prediction accuracy and robustness. The experimental results show that the prediction error MSE = 5.83, RMSE = 2.41, and MAPE = 42.37% of the PLS-LSTM model is significantly lower than that of the traditional model, which provides a new idea for rainfall prediction and technical support for the practical applications in meteorological warning and related fields.