Rainfall forecasting has earned significant studies importance in popular years due to its complexities and rainfall prediction is currently more challenging than ever due to significant changes in the climate. By extracting hidden patterns from previous weather observations, machine learning techniques can predict rainfall.Previous models implement complex statistical models that are often too expensive, both numerically and financially, or inappropriate for downstream applications. In this study of predicting future hourly rainfall high volume using time-series data, models based on Stacked LSTM, XGBoost, and an ensemble, Radial Basis Function Networks (RBFNs), were compared. Climate datasets rom five major cities in the UK cities were used from 2000 to 2020. The models' effectiveness was evaluated using the accuracy, Precision, recall, analysis metrics Loss, the root mean squared error, R-square values,Mean Absolute Error, and Root Mean Squared Logarithmic Anomaly. The predicted resultsshow tetter performance comparing Radial Basis Function Networks (RBFNs) and Stacked-LSTM Networks as rainfall prediction models. The conclusion suggests that models for based on Radial Basis Function Networks (RBFNs) with fewer concealed layers, our approach performs better, highlighting it's appropriate for budget-wise rainfall forecasting applications.

Modern Deep Learning, Accuracy, Precision, Recall, RBFNs, LSTM, MAE, RMSE, RAE, Pytorch, SARIMA.