Time Series Classification

시계열 데이터 분류

1. Without data representation

• DataFrame 형태의 시계열 데이터를 입력으로 활용하는 time series classification에 대한 설명.
• 입력 데이터 형태 : TXP (P>=2) 차원의 다변량 시계열 데이터(multivariate time-series data)

time series classification 사용 시, 설정해야하는 값

• 시계열 분류 모델 :

  • LSTM
  • GRU
  • 1D CNN

• 시계열 분류 모델 hyperparameter : 아래에 자세히 설명.

  • LSTM hyperparameter
  • GRU hyperparameter
  • 1D CNN hyperparameter

python time series classification.py --model='lstm' \
                                     --attention=False \
                                     --hidden_size=20 \
                                     --num_layers=2 \
                                     --dropout=0.1 \
                                     --bidirectional=False \

시계열 분류 모델 hyperparameter

1. LSTM & GRU

• attention : If True, adds an attention layer to RNN. Default: False
• hidden_size : The number of features in the hidden state h
• num_layers : The number of recurrent layers. E.g., setting num_layers=2 would mean stacking two RNNs together to form a stacked RNN, with the second RNN taking in outputs of the first RNN and computing the final results. Default: 1
• dropout : If non-zero, introduces a Dropout layer on the outputs of each RNN layer except the last layer, with dropout probability equal to dropout. Default: 0
• bidirectional : If True, becomes a bidirectional RNN. Default: False
• bias : If False, then the layer does not use bias weights b_ih and b_hh. Default: True

2. 1D CNN

• num_layers : Number of convolutional layers.
• activation : Type of activation functions to be used. Default : relu
• dropout : If non-zero, introduces a Dropout layer on the outputs of each CNN layer except the last layer, with dropout probability equal to dropout. Default: 0
• batch_norm : If True, applies Batch Normalization after CNN layers. Default: False
• kernel_size : Size of the convolving kernel
• stride : Stride of the convolution. Default: 1
• padding : Padding added to both sides of the input. Default: 0
• dilation : Spacing between kernel elements. Default: 1
• bias : If True, adds a learnable bias to the output. Default: True

2. With data representation

• 일정한 형식의 representation을 입력으로 활용하는 classification에 대한 설명.
• 입력 데이터 형태 : P (P>=2) 차원 벡터
  

python time series classification with data representation.py --model='fc' \
                                                              --num_layers=2 \
                                                              --activation=relu \
                                                              --dropout=0.2 \
                                                              --batch_norm=True

time series classification 사용 시, 설정해야하는 값

• 분류 모델 :

  • FC layers (Fully Connected layers)

• 분류 모델 hyperparameter : 아래에 자세히 설명.

  • FC layers (Fully Connected layers)

분류 모델 hyperparameter

1. FC layers

• num_layers : The number of linear layers.
• activation : Type of activation functions to be used. Default : relu
• dropout : If non-zero, introduces a Dropout layer on the outputs of each CNN layer except the last layer, with dropout probability equal to dropout. Default: 0
• batch_norm : If True, applies Batch Normalization after CNN layers. Default: False
• bias : If True, adds a learnable bias to the output. Default: True