The evapotranspiration rate can be used to estimate water loss. However, there are 31 equations available to be chosen, and randomly choosing the equation might not project the actual results. This is very crucial because, without the equation, we cannot proceed with the parameter selection. These findings can justify the parameter chosen for the prediction model development. Long Short-Term Memory (LSTM) is known for its ability to retain memory better than Recurrent Neural Network (RNN). This is due to LSTM architecture, where the memory cell is available to store memory for long-term dependency. RNN suffers from a vanishing gradient that can affect the prediction, whether in accuracy, precision, etc. LSTM was developed specifically to address the issue of RNN. Even though LSTM is better overall, it can be further enhanced. The proposed method is to adjust the Hyperparameter Settings and combine them with Hybridization. Our findings indicate that the prediction accuracy improved significantly. The hybrid model chosen was Gated Recurrent Unit (GRU), combined with LSTM and Hyperparameter Settings, resulting in the best and highest prediction accuracy compared to the LSTM Vanilla and LSTM with Hyperparameter Settings. LSTM Hyperparameter Settings and Hybridization dominate the top three scores. The scoring stretched until 11th place before the LSTM Hyperparameter Settings score came in. The top three scores were for Case 99, Case 36, and Case 90 with 0.0626, 0.06446, 0.06606 MAE, 0.00667, 0.00706, 0.00759 MSE, 0.0817, 0.084, 0.0871 RMSE and 0.99261, 0.99219, 0.9916 R , respectively. As for the LSTM Hyperparameter Settings score, 0.0712 MAE, 0.00861 MSE, 0.09278 RMSE, and 0.99047 R . © 2022 Little Lion Scientific.