keras-anomaly-detection 代码分析——本质上就是SAE、LSTM时间序列预测

keras-anomaly-detection

Anomaly detection implemented in Keras

The source codes of the recurrent, convolutional and feedforward networks auto-encoders for anomaly detection can be found in keras_anomaly_detection/library/convolutional.py and keras_anomaly_detection/library/recurrent.py and keras_anomaly_detection/library/feedforward.py

The the anomaly detection is implemented using auto-encoder with convolutional, feedforward, and recurrent networks and can be applied to:

• timeseries data to detect timeseries time windows that have anomaly pattern
  • LstmAutoEncoder in keras_anomaly_detection/library/recurrent.py
  • Conv1DAutoEncoder in keras_anomaly_detection/library/convolutional.py
  • CnnLstmAutoEncoder in keras_anomaly_detection/library/recurrent.py
  • BidirectionalLstmAutoEncoder in keras_anomaly_detection/library/recurrent.py
• structured data (i.e., tabular data) to detect anomaly in data records
  • Conv1DAutoEncoder in keras_anomaly_detection/library/convolutional.py
  • FeedforwardAutoEncoder in keras_anomaly_detection/library/feedforward.py
     
    看LSTM的模型吧：

    ?

    1 2 3 4 5 6 7 8 9

    def create_model(time_window_size, metric):         model = Sequential()         model.add(LSTM(units=128, input_shape=(time_window_size, 1), return_sequences=False))           model.add(Dense(units=time_window_size, activation='linear'))           model.compile(optimizer='adam', loss='mean_squared_error', metrics=[metric])         print(model.summary()) return model

    再看feedforward的模型：

    ?

    1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

    def create_model(self, input_dim):         encoding_dim = 14         input_layer = Input(shape=(input_dim,))           encoder = Dense(encoding_dim, activation="tanh",                         activity_regularizer=regularizers.l1(10e-5))(input_layer)         encoder = Dense(encoding_dim // 2, activation="relu")(encoder)           decoder = Dense(encoding_dim // 2, activation='tanh')(encoder)         decoder = Dense(input_dim, activation='relu')(decoder)           model = Model(inputs=input_layer, outputs=decoder)         model.compile(optimizer='adam',                       loss='mean_squared_error', metrics=['accuracy'])

     CNN的：

    ?

    1 2 3 4 5 6 7 8 9 10 11

    def create_model(time_window_size, metric):     model = Sequential()     model.add(Conv1D(filters=256, kernel_size=5, padding='same', activation='relu',                      input_shape=(time_window_size, 1)))     model.add(GlobalMaxPool1D())       model.add(Dense(units=time_window_size, activation='linear'))       model.compile(optimizer='adam', loss='mean_squared_error', metrics=[metric])     print(model.summary())     return model

     都是将输出设置成自己，异常点就是查看偏离那90%的预测error较大的点。