Rainfall Forecasting for Silchar City using Stacked- LSTM

Accurate forecasting of rainfall is a very difficult task in meteorology. The complexity, execution time and high computing power required by the Numerical Weather Prediction (NWP) models have always motivated researchers to develop alternate rainfall prediction models. As weather data is a time series by nature this paper proposes a Stacked Long Short-Term Memory (Stacked-LSTM) based Recurrent Neural Network for rainfall prediction of Silchar city in north-east India. To further increase the efficiency of the model Boruta feature selection is used. The proposed model has been compared with LSTM, XGBoost, Random Forest and Multiple Linear Regression. The performance measures used in this study are Root Mean Squared Error (RMSE) and R-Squared values. The proposed model widely outperforms all other models with an RMSE value of 0.98 and R-Squared value of 97.03%.