SecuritAI LSTM RNN Model

securitai-lstm-model.py

model = Sequential()
model.add(Embedding(num_words, 32, input_length=max_log_length))
# Prevent overfitting using dropout method of regularization
model.add(Dropout(0.5))
model.add(LSTM(64, recurrent_dropout=0.5))
model.add(Dropout(0.5))
# Condense to single binary output value
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
# Training set automatically split 75/25 to check validation loss/accuracy at each epoch
model.fit(X_train, Y_train, validation_split=0.25, epochs=3, batch_size=128, callbacks=[tb_callback])
# Evaluation of separate test dataset performed after training
score, acc = model.evaluate(X_test, Y_test, verbose=1, batch_size=128)