BTCUSD coinbase_30m

• 지표로 통해 사용될수 있는 feature는 총 68개 time을 빼면
• 결측치(NaN) 데이터 Drop 하여 모든 지표 값 가용가능하게 함

• 결측치를 제외
  • (20700, 69) → (20501, 40)
  • 사용가능한 feature 40 개 → input_dim

Data Scaler

#   스케일러 설정 
"""
    open, high, low, close, Mean, Upper, Lower, Upper fib,
    low fib, Plot, Plot.1, Plot.2, ParabolicSAR, H, M, L, 
    RES, SUP, EMA_34, short_line, VWAP, 34 EMA, hlc3
"""
price_scaler = MinMaxScaler()
""" Volume.1, A/D Volume """
vol_scaler = MinMaxScaler()
""" OnBalanceVolume """
obv_scaler = MinMaxScaler()
""" MF """
mf_scaler = MinMaxScaler()
""" Accumulation/Distribution """
ad_scaler = MinMaxScaler()
""" RSI """
rsi_scaler = MinMaxScaler()
""" Second RSI """
rsi2_scaler = MinMaxScaler()
""" Second MFI """
mfi_scaler = MinMaxScaler()
""" BB Basis, BB Upper, BB Lower """
bb_scaler = MinMaxScaler()
""" Plot.3, Plot.4, CCI """
cci_scaler = MinMaxScaler()

#   스케일 데이터 fitting
price_scaler.fit(np.array(copy_data[['open', 'high', 'low', 'close', 'Mean ', 'Upper ', 
                                     'Lower', 'Upper fib ', 'low fib', 'Plot', 'Plot.1', 
                                     'Plot.2', 'ParabolicSAR', 'H', 'M', 'L', 'RES', 'SUP', 
                                     'EMA_34', 'short_line', 'VWAP', '34 EMA', 'hlc3']]).reshape(-1, 1))
vol_scaler.fit(np.array(copy_data[['Volume.1', 'A/D Volume']]).reshape(-1, 1))
obv_scaler.fit(np.array(copy_data[['OnBalanceVolume']]).reshape(-1, 1))
mf_scaler.fit(np.array(copy_data[['MF']]).reshape(-1, 1))
ad_scaler.fit(np.array(copy_data[['Accumulation/Distribution']]).reshape(-1, 1))
rsi_scaler.fit(np.array(copy_data[['RSI']]).reshape(-1, 1))
rsi2_scaler.fit(np.array(copy_data[['Second RSI']]).reshape(-1, 1))
mfi_scaler.fit(np.array(copy_data[['Second MFI']]).reshape(-1, 1))
bb_scaler.fit(np.array(copy_data[['BB Basis', 'BB Upper', 'BB Lower']]).reshape(-1, 1))
cci_scaler.fit(np.array(copy_data[['Plot.3', 'Plot.4', 'CCI']]).reshape(-1, 1))

#   학습데이터 0 ~ 1 스케일 
copy_data[['open', 'high', 'low', 'close', 'Mean ', 'Upper ', 
           'Lower', 'Upper fib ', 'low fib', 'Plot', 'Plot.1', 
           'Plot.2', 'ParabolicSAR', 'H', 'M', 'L', 'RES', 'SUP', 
           'EMA_34', 'short_line', 'VWAP', '34 EMA', 'hlc3']] = price_scaler.transform(np.array(copy_data[['open', 'high', 'low', 'close', 'Mean ', 
                                                                                                   'Upper ', 'Lower', 'Upper fib ', 'low fib', 
                                                                                                   'Plot', 'Plot.1', 'Plot.2', 'ParabolicSAR', 
                                                                                                   'H', 'M', 'L', 'RES', 'SUP', 'EMA_34', 'short_line', 
                                                                                                   'VWAP', '34 EMA', 'hlc3']]).reshape(-1, 23))
copy_data[['Volume.1', 'A/D Volume']] = vol_scaler.transform(np.array(copy_data[['Volume.1', 'A/D Volume']]).reshape(-1, 2))
copy_data[['OnBalanceVolume']] = obv_scaler.transform(np.array(copy_data[['OnBalanceVolume']]).reshape(-1, 1))
copy_data[['MF']] = mf_scaler.transform(np.array(copy_data[['MF']]).reshape(-1, 1))
copy_data[['Accumulation/Distribution']] = ad_scaler.transform(np.array(copy_data[['Accumulation/Distribution']]).reshape(-1, 1))
copy_data[['RSI']] = rsi_scaler.transform(np.array(copy_data[['RSI']]).reshape(-1, 1))
copy_data[['Second RSI']] = rsi2_scaler.transform(np.array(copy_data[['Second RSI']]).reshape(-1, 1))
copy_data[['Second MFI']] = mfi_scaler.transform(np.array(copy_data[['Second MFI']]).reshape(-1, 1))
copy_data[['BB Basis', 'BB Upper', 'BB Lower']] = bb_scaler.transform(np.array(copy_data[['BB Basis', 'BB Upper', 'BB Lower']]).reshape(-1, 3))
copy_data[['Plot.3', 'Plot.4', 'CCI']] = cci_scaler.transform(np.array(copy_data[['Plot.3', 'Plot.4', 'CCI']]).reshape(-1, 3))

Scaled Chart Data

Data Corr