Improving the Forecasting Accuracy Based on the Lunar Calendar in Modeling Rainfall Levels Using the Bi-LSTM Method through the Grid Search Approach

Table 4

Results of Bi-LSTM-grid search tuning parameters on the Gregorian calendar.

Forecasting length	Parameter				MSE
	Neuron	Batch	Epoch	Dropout
3	20	4	200	0.1	0.01926
6	20	4	200	0.1	0.01945
12	20	4	200	0.2	0.01882
18	15	4	200	0.1	0.02004
24	20	4	200	0.1	0.01964