Extreme precipitation events pose significant challenges to societal infrastructure and environmental stability, particularly in vulnerable regions such as the Beijing-Tianjin-Hebei area of China. Traditional forecasting methods frequently fail to capture the complex nonlinear dynamics of such events. This study proposes a novel hybrid model, variational mode decomposition-principal component analysis-extreme gradient boosting (VMD-PCA-XGBoost), integrating VMD for signal processing, PCA for dimensionality reduction, and XGBoost for predictive accuracy. Using 2023 data from 13 global navigation satellite system and meteorological stations, the model is rigorously compared with XGBoost and empirical-mode-decomposition-based models. It achieves average critical success index, probability of detection, and false alarm rate values of 52.14%, 73.07%, and 35.98%, respectively. These findings demonstrate the model's robustness and precision and provide a promising tool for improving precipitation forecasts and disaster preparedness in climate-sensitive regions.