def azureml_main(frame1):
# Set backend
import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
import milkutilities as mu
from sklearn import linear_model
Azure = False
X = frame1[['Month.Count', 'monthNumCubed']].as_matrix()
Y = frame1['Milk.Prod'].as_matrix()
regr = linear_model.LinearRegression()
regr.fit(X, Y)
frame1['Predicted'] = regr.predict(X).tolist()
frame1['Resid'] = frame1['Milk.Prod'] - frame1['Predicted']
## Add a time index to the data frame
frame1 = mu.add_time_index(frame1)
fig1 = plt.figure(1, figsize = (12,9))
ax = fig1.gca()
frame1[['Milk.Prod', 'Predicted']].plot(ax = ax)
plt.xlabel("Date")
plt.ylabel("Log CA milk production")
plt.title("Log of milk produciton vs. date")
plt.show()
if(Azure == True): fig1.savefig('scatter1.png')
fig2 = plt.figure(1, figsize = (12,9))
fig2.clf()
ax = fig2.gca()
frame1['Resid'].plot(ax = ax)
plt.xlabel("Date")
plt.ylabel("Residuals of linear model")
plt.title("Residuals of linear model vs. date")
plt.show()
if(Azure == True): fig2.savefig('scatter2.png')
fig3 = plt.figure(1, figsize = (12,9))
fig3.clf()
ax = fig3.gca()
frame1.boxplot( column = ['Resid'], ax = ax,
by = ['Month.Number','Month'])
plt.xlabel("Month of year")
plt.ylabel("Residuals of linear model")
plt.title("Residuals of linear model by Month")
plt.show()
if(Azure == True): fig3.savefig('scatter3.png')
return frame1
def azureml_main(frame1):
# Set backend
import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
import milkutilities as mu
from sklearn import linear_model
Azure = False
X = frame1[['Month.Count', 'monthNumCubed']].as_matrix()
Y = frame1['Milk.Prod'].as_matrix()
regr = linear_model.LinearRegression()
regr.fit(X, Y)
frame1['Predicted'] = regr.predict(X).tolist()
frame1['Resid'] = frame1['Milk.Prod'] - frame1['Predicted']
## Add a time index to the data frame
frame1 = mu.add_time_index(frame1)
fig1 = plt.figure(1, figsize=(12, 9))
ax = fig1.gca()
frame1[['Milk.Prod', 'Predicted']].plot(ax=ax)
plt.xlabel("Date")
plt.ylabel("Log CA milk production")
plt.title("Log of milk produciton vs. date")
plt.show()
if (Azure == True): fig1.savefig('scatter1.png')
fig2 = plt.figure(1, figsize=(12, 9))
fig2.clf()
ax = fig2.gca()
frame1['Resid'].plot(ax=ax)
plt.xlabel("Date")
plt.ylabel("Residuals of linear model")
plt.title("Residuals of linear model vs. date")
plt.show()
if (Azure == True): fig2.savefig('scatter2.png')
fig3 = plt.figure(1, figsize=(12, 9))
fig3.clf()
ax = fig3.gca()
frame1.boxplot(column=['Resid'], ax=ax, by=['Month.Number', 'Month'])
plt.xlabel("Month of year")
plt.ylabel("Residuals of linear model")
plt.title("Residuals of linear model by Month")
plt.show()
if (Azure == True): fig3.savefig('scatter3.png')
return frame1
def azureml_main(frame1):
import pandas as pd
import numpy as np
import milkutilities as mu
import os.path
## If not in the Azure environment, read the data from a csv
## file for testing purposes. Use os.path.join to create an
## OS independent way to create the full path, which should
## work on Windows, Mac, and Linux.
Azure = False
if(Azure == False):
pathName = "C:/Users/Steve/Documents/AzureML/Data Sets/CA_Milk"
fileName = "cadairydata.csv"
filePath = os.path.join(pathName, fileName)
frame1 = pd.read_csv(filePath)
## Trim the month codes to 3 characters to ensure they
## are sonsistent.
frame1['Month'] = frame1['Month'].map(lambda x: str(x)[:3])
## Add a time index to the data frame
frame1 = mu.add_time_index(frame1)
## Add a date-time type column.
# frame1['datetime'] = frame1['Year'].apply(str) + '-' + frame1['Month.Number'].apply(str)+ '-01'
# frame1['datetime'] = pd.to_datetime(frame1['datetime'], format="%Y-%m-%d")
## Compute new columns containing the polynomial values
## of the count of months.
frame1['Month.Count'] = frame1['Month.Number'] + \
12 * (frame1['Year'] - 1995)
x = frame1['Month.Count'].as_matrix()
frame1['monthNumSqred'] = np.power(x, 2).tolist()
frame1['monthNumCubed'] = np.power(x, 3).tolist()
return frame1
def azureml_main(frame1):
import pandas as pd
import milkutilities as mu
import os.path
## If not in the Azure environment, read the data from a csv
## file for testing purposes. Use os.path.join to create an
## OS independent way to create the full path, which should
## work on Windows, Mac, and Linux.
Azure = False
if(Azure == False):
pathName = "C:/Users/Steve/Documents/AzureML/Data Sets/CA_Milk"
fileName = "cadairydata.csv"
filePath = os.path.join(pathName, fileName)
frame1 = pd.read_csv(filePath)
## Add a time index to the data frame
frame1 = mu.add_time_index(frame1)
## Cut the last 12 months off the end of the
## data frame using pandas time series indexing.
frame2 = frame1[:'2013-01-01']
return frame2
def azureml_main(frame1):
# Set backend
import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
import milkutilities as mu
import statsmodels.api as sm
import pandas as pd
Azure = False
## Add a time index to the data frame
frame1 = mu.add_time_index(frame1)
## Compute the residuals
frame1['Resid'] = frame1['Milk.Prod'] - frame1['Scored Labels']
fig1 = plt.figure(1, figsize=(12, 9))
ax = fig1.gca()
frame1[['Milk.Prod', 'Scored Labels']].plot(ax=ax)
plt.xlabel("Date")
plt.ylabel("Log CA milk production")
plt.title("Log of milk produciton vs. date")
plt.show()
if (Azure == True): fig1.savefig('scatter1.png')
fig2 = plt.figure(1, figsize=(12, 9))
fig2.clf()
ax = fig2.gca()
frame1['Resid'].plot(ax=ax)
plt.xlabel("Date")
plt.ylabel("Residuals of linear model")
plt.title("Residuals of linear model vs. date")
plt.show()
if (Azure == True): fig2.savefig('scatter2.png')
fig3 = plt.figure(1, figsize=(12, 9))
fig3.clf()
ax = fig3.gca()
frame1.boxplot(column=['Resid'], ax=ax, by=['Month.Number', 'Month'])
plt.xlabel("Month of year")
plt.ylabel("Residuals of linear model")
plt.title("Residuals of linear model by Month")
plt.show()
if (Azure == True): fig3.savefig('scatter3.png')
## QQ Normal plot of residuals
fig4 = plt.figure(figsize=(12, 6))
fig4.clf()
ax = fig4.gca()
sm.qqplot(frame1['Resid'], ax=ax)
ax.set_title('QQ Normal plot of residuals')
if (Azure == True): fig4.savefig('plot4.png')
## Histograms of the residuals
fig5 = plt.figure(figsize=(12, 6))
fig5.clf()
fig5.clf()
ax = fig5.gca()
ax.hist(frame1['Resid'].as_matrix(), bins=40)
ax.set_xlabel("Model residuals")
ax.set_ylabel("Density")
ax.set_title("Histogram of residuals")
if (Azure == True): fig5.savefig('plot5.png')
## Compute a data frame with the rms errors for the model
out_frame = pd.DataFrame({ \
'rmse_Overall' : [rmse(frame1['Resid'])], \
'rmse_test' : [rmse(frame1.ix['2013-01-31':, 'Resid'])] })
return out_frame
def azureml_main(frame1):
# Set backend
import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
import milkutilities as mu
import statsmodels.api as sm
import pandas as pd
Azure = False
## Add a time index to the data frame
frame1 = mu.add_time_index(frame1)
## Compute the residuals
frame1['Resid'] = frame1['Milk.Prod'] - frame1['Scored Labels']
fig1 = plt.figure(1, figsize = (12,9))
ax = fig1.gca()
frame1[['Milk.Prod', 'Scored Labels']].plot(ax = ax)
plt.xlabel("Date")
plt.ylabel("Log CA milk production")
plt.title("Log of milk produciton vs. date")
plt.show()
if(Azure == True): fig1.savefig('scatter1.png')
fig2 = plt.figure(1, figsize = (12,9))
fig2.clf()
ax = fig2.gca()
frame1['Resid'].plot(ax = ax)
plt.xlabel("Date")
plt.ylabel("Residuals of linear model")
plt.title("Residuals of linear model vs. date")
plt.show()
if(Azure == True): fig2.savefig('scatter2.png')
fig3 = plt.figure(1, figsize = (12,9))
fig3.clf()
ax = fig3.gca()
frame1.boxplot( column = ['Resid'], ax = ax,
by = ['Month.Number','Month'])
plt.xlabel("Month of year")
plt.ylabel("Residuals of linear model")
plt.title("Residuals of linear model by Month")
plt.show()
if(Azure == True): fig3.savefig('scatter3.png')
## QQ Normal plot of residuals
fig4 = plt.figure(figsize = (12,6))
fig4.clf()
ax = fig4.gca()
sm.qqplot(frame1['Resid'], ax = ax)
ax.set_title('QQ Normal plot of residuals')
if(Azure == True): fig4.savefig('plot4.png')
## Histograms of the residuals
fig5 = plt.figure(figsize = (12,6))
fig5.clf()
fig5.clf()
ax = fig5.gca()
ax.hist(frame1['Resid'].as_matrix(), bins = 40)
ax.set_xlabel("Model residuals")
ax.set_ylabel("Density")
ax.set_title("Histogram of residuals")
if(Azure == True): fig5.savefig('plot5.png')
## Compute a data frame with the rms errors for the model
out_frame = pd.DataFrame({ \
'rmse_Overall' : [rmse(frame1['Resid'])], \
'rmse_test' : [rmse(frame1.ix['2013-01-31':, 'Resid'])] })
return out_frame