Regression+LSTM.py

#%%
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.layers import LSTM
from tensorflow.keras.layers import Dropout
from tensorflow.keras.models import load_model

from sklearn.linear_model import LinearRegression
from sklearn.preprocessing import MinMaxScaler
import joblib

import numpy as np
import pandas as pd
import os

#設定LSTM往前看的筆數和預測筆數
LookBackNum = 12 #LSTM往前看的筆數
ForecastNum = 48 #預測筆數

#載入訓練資料
DataName = os.getcwd()+'\ExampleTrainData(AVG)\AvgDATA_17.csv'
SourceData = pd.read_csv(DataName, encoding='utf-8')

#迴歸分析 選擇要留下來的資料欄位
#(溫度,濕度,光照度)
#(發電量)
Regression_X_train = SourceData[['Temperature(°C)','Humidity(%)','Sunlight(Lux)']].values
Regression_y_train = SourceData[['Power(mW)']].values

#LSTM 選擇要留下來的資料欄位
#(溫度,濕度,光照度)
AllOutPut = SourceData[['Temperature(°C)','Humidity(%)','Sunlight(Lux)']].values

#正規化
LSTM_MinMaxModel = MinMaxScaler().fit(AllOutPut)
AllOutPut_MinMax = LSTM_MinMaxModel.transform(AllOutPut)

X_train = []
y_train = []

#設定每i-12筆資料(X_train)就對應到第i筆資料(y_train)
for i in range(LookBackNum,len(AllOutPut_MinMax)):
  X_train.append(AllOutPut_MinMax[i-LookBackNum:i, :])
  y_train.append(AllOutPut_MinMax[i, :])


X_train = np.array(X_train)
y_train = np.array(y_train)

# Reshaping
#(samples 是訓練樣本數量,timesteps 是每個樣本的時間步長,features 是每個時間步的特徵數量)
X_train = np.reshape(X_train,(X_train.shape [0], X_train.shape [1], 3))


#%%
#============================建置&訓練「LSTM模型」============================
#建置LSTM模型

regressor = Sequential ()

regressor.add(LSTM(units = 128, return_sequences = True, input_shape = (X_train.shape[1], 3)))

regressor.add(LSTM(units = 64, return_sequences = True))

regressor.add(Dropout(0.2))

regressor.add(LSTM(units = 16 ))

regressor.add(Dropout(0.2))

# output layer
regressor.add(Dense(units = 3))
regressor.compile(optimizer = 'adam', loss = 'mean_squared_error')

#開始訓練
regressor.fit(X_train, y_train, epochs = 100, batch_size = 128)

#保存模型
from datetime import datetime
NowDateTime = datetime.now().strftime("%Y-%m-%dT%H_%M_%SZ")
regressor.save('WheatherLSTM_'+NowDateTime+'.h5')
print('Model Saved')



#%%
#============================建置&訓練「回歸模型」========================

#開始迴歸分析(對發電量做迴歸)
RegressionModel = LinearRegression()
RegressionModel.fit(LSTM_MinMaxModel.transform(Regression_X_train), Regression_y_train)

#儲存回歸模型
from datetime import datetime
NowDateTime = datetime.now().strftime("%Y-%m-%dT%H_%M_%SZ")
joblib.dump(RegressionModel, 'WheatherRegression_'+NowDateTime)

#取得截距
print('截距: ',RegressionModel.intercept_)

#取得係數
print('係數 : ', RegressionModel.coef_)

#取得R平方
print('R squared: ',RegressionModel.score(LSTM_MinMaxModel.transform(Regression_X_train), Regression_y_train))