Skip to main content
. 2023 Jun 16;2023:330–339.

Figure 2.

Figure 2.

(A) Architecture of the hourly LSTM model. We trained this model using the dataset aggregated by hour. The top architecture included 2 LSTM layers with a 0.5 dropout layer between the final LSTM layer and linear fully connected layer. (B) Architecture of the daily LSTM model. We trained this model using the dataset aggregated by day. The top architecture included 2 LSTM layers and a linear fully connected layer. LSTM: long short-term memory.