In order to improve the accuracy and efficiency of rainfall forecasting, a network based on bi-directional long short-term memory (BiLSTM) and rat swarm and jellyfish search optimizer (RSJSO) is proposed. The rainfall data is preprocessed by using missing data interpolation and feature fusion technology, and the feature extraction process is optimized by chord distance and based on rider optimization algorithm and based neural network (RideNN). Oversampling techniques are used to enhance the dataset. The experimental results show that the model performs well in a variety of evaluation indicators, and can provide reliable support for the relevant decision-making of refined rainfall information, and effectively enhance the ability to respond to emergencies in complex scenarios.