Rainfall forecasting remains one of the primary concerns for meteorologists, given its crucial role in areas such as agriculture, water management, ecosystems, hydrology, and even the economy. However, with ongoing climate change and increasing human impact on the environment, predicting rainfall patterns has become significantly more challenging. In response to this, we propose a novel and reliable hybrid model for rainfall prediction: the BPNN-PSO (Back Propagation Neural Network with Particle Swarm Optimization), offering a timely solution to these difficulties. BPNN and BPNN-PSO algorithms were used to analyze 1974–2022 monthly meteorological data at Qazigund station in the Jhelum River basin, India. The result showed that compared with classical BPNN uses the correlation coefficient of BPNN-PSO better. The accuracy of estimated rainfalls based on BPNN-PSO can be improved significantly compared to a single BPNN. The RMSE (Root Mean Square Error) and MSE help assess how accurately a prediction model mimics actual observations, while NSE and R2 represent proportion variance accounted for by independent variables. The BPNN-PSO is a reliable and appropriate way of forecasting rainfall; it can be used for worldwide climate forecasts, with high-volume data processing needed.