Comedor Temperature Prediction--Blanco-Córdoba-Sora

A 2018-2019/Comedor Temperature Prediction--Blanco-Cordoba-Sora/Comedor Temperature Prediction.py

@@ -0,0 +1,171 @@
+# -*- coding: utf-8 -*-
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import os
+
+
+#Importing the external files
+
+WhereToSaveFiles= r"C:\Users\Giulia\Desktop\Comedor Temperature Prediction--Blanco-Cordoba-Sora"
+os.chdir(WhereToSaveFiles)
+ExternalFilesFolder = r"C:\Users\Giulia\Desktop\Comedor Temperature Prediction--Blanco-Cordoba-Sora"
+DataFileName="SpanishDataSet.xlsx"
+path_DataFileName=os.path.join(ExternalFilesFolder,DataFileName)
+
+DF_DataFile=pd.read_excel(path_DataFileName,index_col=0,header=1)
+DF_DataFile.columns
+
+
+#We transform the data in the file from 'undefined objects' to 'time and date data'
+
+PreviousIndex=DF_DataFile.index
+NewParsedIndex=pd.to_datetime(PreviousIndex)
+DF_DataFile.index=NewParsedIndex
+
+DF_DataFile.head(4)
+DF_DataFile.tail(4)
+
+DF_DataFile_cleaned=DF_DataFile.dropna()
+DF_DataFile_cleaned.corr()
+
+
+#We extract/select the data we are interested in:
+
+DF_TemperatureInsideComedor=DF_DataFile["3:Temperature_Comedor_Sensor"]["2012-03-13 11:45:00":"2012-04-11 06:30:00"] #all the possible data
+
+
+#We plot the TemperatureInsideComedor measured versus time during the period we are considering.
+
+plt.figure()
+DF_TemperatureInsideComedor.plot()
+plt.xlabel("Time")
+plt.ylabel("Temperature measured in the Comedor")
+plt.show()
+
+
+#Now we import the data in order to create our new dataframe
+
+DF_TemperatureInsideComedor=DF_DataFile[["3:Temperature_Comedor_Sensor"]]
+DF_TemperatureInsideComedor.head(10)
+
+DF_TemperatureInsideHabitacion=DF_DataFile[["4:Temperature_Habitacion_Sensor"]]
+DF_TemperatureInsideHabitacion.head(10)
+
+DF_WeatherTemperature=DF_DataFile[["5:Weather_Temperature"]]
+DF_WeatherTemperature.head(10)
+
+DF_TemperatureExterior=DF_DataFile[["22:Temperature_Exterior_Sensor"]]
+DF_TemperatureExterior.head(10)
+
+
+DF_DataComedor_joined=DF_TemperatureInsideComedor.join([DF_TemperatureInsideHabitacion,DF_WeatherTemperature,DF_TemperatureExterior])
+DF_DataComedor_joined.columns=["TempInsCom","TempInsHab", "WeaTemp", "TempExt"]
+DF_DataComedor_joined.to_excel("DataComedor_joinedOk.xlsx")
+
+
+#Doing the lag
+
+DF_modLag=DF_DataComedor_joined.copy()
+DF_modLag.dropna(inplace=True)
+df= DF_modLag
+
+
+def lag_feature(df,column_name, lag_start,lag_end,lag_interval):
+    """This function takes add inputs the name of the file, the column name on which we want to do the shift,
+        the first step, the last step, and the magnitude of the single step. It returns the file with each columns shifted as requested"""
+    for i in range(lag_start,lag_end+1, lag_interval): 
+        new_column_name= column_name+ "-15X" + str(i) + "min"  
+        print new_column_name
+        df[new_column_name]=df[column_name].shift(i)
+        df.dropna(inplace=True)
+    return df
+
+
+DF_modLag= lag_feature(df, "TempInsHab",1,5,1)
+DF_modLag.head()
+DF_modLag= lag_feature(df, "WeaTemp",1,5,1)
+DF_modLag.head()
+DF_modLag= lag_feature(df, "TempExt",1,5,1)
+DF_modLag.head()
+
+DF_modLag.corr()
+
+
+DF_modLag.to_excel("DF_FinalDataSetokok.xlsx")
+FinalDataSetToPredict="DF_FinalDataSetokok.xlsx"
+path_DataToPredict=os.path.join(ExternalFilesFolder,FinalDataSetToPredict)
+DF_FinalDataSetTopredict=pd.read_excel(path_DataToPredict,index_col=0,header=0)
+
+
+#First Normalizing, before doing my machine learning
+
+DF_modLagChosenDatas=DF_FinalDataSetTopredict["2012-03-13":"2012-04-11"]
+DF_modLagChosenDatas.head()
+DF_modLagChosenDatas.dropna()
+
+
+def normalizing(DF_modLagChosenDatas):
+    """This function normalize the data.It calculates the minimum and the maximum values for each columns, then it does the normalization"""
+    maximum=DF_modLagChosenDatas.max()
+    minimum=DF_modLagChosenDatas.min()
+    return (DF_modLagChosenDatas-minimum)/(maximum-minimum)
+
+#Now after making my Final Data Set, we are going to start our Machine Learning
+
+DF_target=DF_modLagChosenDatas["TempInsCom"]
+DF_features=DF_modLagChosenDatas.drop("TempInsCom",axis=1)
+
+#Normalizing
+
+DF_FinalDataSet_ChosenDatas_norm = normalizing(DF_modLagChosenDatas)
+DF_FinalDataSet_ChosenDatas_norm.corr()
+DF_target_norm = DF_FinalDataSet_ChosenDatas_norm["TempInsCom"]
+DF_features_norm = DF_FinalDataSet_ChosenDatas_norm.drop("TempInsCom",axis=1)
+
+#Now we divide the data in two groups: the one on which we do the regression and the one on which we test the regression
+
+from sklearn.model_selection import train_test_split
+X_train,X_test,Y_train,Y_test=train_test_split(DF_features,DF_target,test_size=0.2,random_state=41234)
+
+
+from sklearn import linear_model
+linear_reg= linear_model.LinearRegression()
+linear_reg.fit(X_train,Y_train)
+predicted_linearreg_split=linear_reg.predict(X_test)
+predicted_DF_linearreg_split=pd.DataFrame(predicted_linearreg_split, index=Y_test.index, columns=["TempInsCom_predicted_linear_reg"])
+predicted_DF_linearreg_split=predicted_DF_linearreg_split.join(Y_test)
+
+#Graph for comparison of the real data and the one predicted from the Linear Regression model
+
+predicted_DF_linearreg_split_alldata=predicted_DF_linearreg_split["2012-03-13":"2012-04-11"]
+predicted_DF_linearreg_split_alldata.plot()
+
+#Now we want to estimate how accurate are our predictions. So, we use the accuracy metrics: MAE, MSE and R2
+
+from sklearn.metrics import mean_absolute_error, mean_squared_error, r2_score
+MAE_linear_reg_split=mean_absolute_error(predicted_linearreg_split,Y_test)
+MSE_linear_reg_split=mean_squared_error(predicted_linearreg_split,Y_test)
+R2_linear_reg_split=r2_score(predicted_linearreg_split,Y_test)
+
+#Let's make other prediction with Cross-Validation model with 40 subsets
+
+from sklearn.model_selection import cross_val_predict
+predicted_linearReg_CV=cross_val_predict(linear_reg,DF_features,DF_target,cv=40)
+predicted_DF_linearReg_CV=pd.DataFrame(predicted_linearReg_CV,index=DF_target.index,columns=["TempInsCom_predicted_linear_reg_CV"])
+predicted_DF_linearReg_CV=predicted_DF_linearReg_CV.join(DF_target)
+
+#Graph for comparison of the real data and the one predicted from the Cross-Validation Model
+
+predicted_DF_linearReg_CV_alldata=predicted_DF_linearReg_CV["2012-03-13":"2012-04-11"]
+predicted_DF_linearReg_CV_alldata.plot()
+
+#We use again the accuracy metrics: MAE, MSE and R2
+
+from sklearn.metrics import mean_absolute_error, mean_squared_error, r2_score
+MAE_linearReg_CV=mean_absolute_error(predicted_linearReg_CV,DF_target)
+MSE_linearReg_CV=mean_squared_error(predicted_linearReg_CV,DF_target)
+R2_linearReg_CV=r2_score(predicted_linearReg_CV,DF_target)
+
+
+