Water Level Prediction of Riam Kanan Dam Using ConvLSTM, BPNN, Gradient Boosting, and XGBoosting Stacking Framework (CLBGXGBoostS)

Abstract

Research focuses on developing a water level prediction framework for the Riam Kanan Dam using an innovative stacking approach called ConvLSTM-BPNN-Gradient Boosting and Stacking XGBoost (CLBGXGBoostS), which combines the strengths of Convolutional Long Short-Term Memory (ConvLSTM), Backpropagation Neural Network (BPNN), and Gradient Boosting. The study aims to evaluate the performance of the CLBGXGBoostS stacking framework in predicting the water level of the Riam Kanan Dam using 5 years of historical data. The results demonstrate that the CLBGXGBoostS framework provides more accurate predictions compared to single models, as evidenced by the Root Mean Squared Error (RMSE) values. CLBGXGBoostS achieves an RMSE of 0.0071, significantly lower than the RMSE of the individual models ConvLSTM (0.1006), BPNN (0.2618), and Gradient Boosting (0.6905). This research contributes to the development of a better water level prediction framework for the Riam Kanan Dam, supporting more effective water resource management and serving as a reference for future research in this field.