Skip to main content
. 2019 Jun 15;21(6):e13803. doi: 10.2196/13803

Figure 2.

Figure 2

Structure of deep learning model. Embedding layer: input_dim is 29,832, which is the size of the dictionary; the input_length for text is 50, input_length for hashtag is 15; the output_dim is 400. Long short-term memory layer: contains 800 units, with dropout=0.2, recurrent_dropout=0.2. Dense layer 1: unit=200, activation=sigmoid. Dense layer 2: unit=200, activation=sigmoid. Dense layer 3: unit=200, activation=sigmoid. Dense layer 4: unit=1, activation=sigmoid. Optimizer: Adam (learning rate set at 0.0001). Loss: Binary_crossentropy. LSTM: long short-term memory.