initial commit

data.csv

+date;quantity
+2015/05/01 00:00:00 UTC;3.0
+2015/05/02 00:00:00 UTC;60.0
+2015/05/03 00:00:00 UTC;59.962507811848994
+2015/05/04 00:00:00 UTC;59.85012495834077
+2015/05/05 00:00:00 UTC;59.663132338081276
+2015/05/06 00:00:00 UTC;59.40199733523725
+2015/05/07 00:00:00 UTC;59.06737265131934
+2015/05/08 00:00:00 UTC;58.66009467376818
+2015/05/09 00:00:00 UTC;58.181181385421375
+2015/05/10 00:00:00 UTC;57.63182982008656
+2015/05/11 00:00:00 UTC;57.013413070580306
+2015/05/12 00:00:00 UTC;56.327476856711186
+2015/05/13 00:00:00 UTC;55.575735661785174
+2015/05/14 00:00:00 UTC;54.760068447290344
+2015/05/15 00:00:00 UTC;53.88251395647168
+2015/05/16 00:00:00 UTC;52.945265618534656
+2015/05/17 00:00:00 UTC;51.95066606621463
+2015/05/18 00:00:00 UTC;50.90120128041495
+2015/05/19 00:00:00 UTC;49.799494376549475
+2015/05/20 00:00:00 UTC;48.64829904811994
+2015/05/21 00:00:00 UTC;47.45049268391651
+2015/05/22 00:00:00 UTC;46.209069176044196
+2015/05/23 00:00:00 UTC;44.92713143675181
+2015/05/24 00:00:00 UTC;43.60788364276732
+2015/05/25 00:00:00 UTC;42.25462322652473
+2015/05/26 00:00:00 UTC;40.87073263430021
+2015/05/27 00:00:00 UTC;39.45967087185806
+2015/05/28 00:00:00 UTC;38.02496485873762
+2015/05/29 00:00:00 UTC;36.57020061279124
+2015/05/30 00:00:00 UTC;35.09901428700723
+2015/05/31 00:00:00 UTC;33.615083081021
+2015/06/01 00:00:00 UTC;32.12211605003108
+2015/06/02 00:00:00 UTC;30.623844834092772
+2015/06/03 00:00:00 UTC;29.124014330961337
+2015/06/04 00:00:00 UTC;27.626373335797986
+2015/06/05 00:00:00 UTC;26.13466517113426
+2015/06/06 00:00:00 UTC;24.65261833051524
+2015/06/07 00:00:00 UTC;23.18393715920739
+2015/06/08 00:00:00 UTC;21.73229259526461
+2015/06/09 00:00:00 UTC;20.301312994094896
+2015/06/10 00:00:00 UTC;18.894575059461395
+2015/06/11 00:00:00 UTC;17.51559490358573
+2015/06/12 00:00:00 UTC;16.167819258698618
+2015/06/13 00:00:00 UTC;14.854616862004274
+2015/06/14 00:00:00 UTC;13.579270035591872
+2015/06/15 00:00:00 UTC;12.344966482339625
+2015/06/16 00:00:00 UTC;11.154791318317827
+2015/06/17 00:00:00 UTC;10.011719361605277
+2015/06/18 00:00:00 UTC;8.91860769679338
+2015/06/19 00:00:00 UTC;7.878188533762637
+2015/06/20 00:00:00 UTC;6.893062378580777
+2015/06/21 00:00:00 UTC;5.965691533591989
+2015/06/22 00:00:00 UTC;5.098393942943337
+2015/06/23 00:00:00 UTC;4.29333739893158
+2015/06/24 00:00:00 UTC;3.552534123651423
+2015/06/25 00:00:00 UTC;2.8778357394881637
+2015/06/26 00:00:00 UTC;2.270928641026093
+2015/06/27 00:00:00 UTC;1.733329779940258
+2015/06/28 00:00:00 UTC;1.266382873407289
+2015/06/29 00:00:00 UTC;0.8712550455122836
+2015/06/30 00:00:00 UTC;0.5489339100463764
+2015/07/01 00:00:00 UTC;0.30022510198663754
+2015/07/02 00:00:00 UTC;0.1257502638281638
+2015/07/03 00:00:00 UTC;0.02594549180161554
+2015/07/04 00:00:00 UTC;0.0010602458597408493
+2015/07/05 00:00:00 UTC;0.05115672615740641
+2015/07/06 00:00:00 UTC;0.17610971758361238
+2015/07/07 00:00:00 UTC;0.3756069027340536
+2015/07/08 00:00:00 UTC;0.6491496425419763
+2015/07/09 00:00:00 UTC;0.9960542226161673
+2015/07/10 00:00:00 UTC;1.4154535621708586
+2015/07/11 00:00:00 UTC;1.9062993812761098
+2015/07/12 00:00:00 UTC;2.4673648210117216
+2015/07/13 00:00:00 UTC;3.09724750997559
+2015/07/14 00:00:00 UTC;3.7943730694818623
+2015/07/15 00:00:00 UTC;4.556999048687759
+2015/07/16 00:00:00 UTC;5.3832192798131775
+2015/07/17 00:00:00 UTC;6.270968642567494
+2015/07/18 00:00:00 UTC;7.218028225874765
+2015/07/19 00:00:00 UTC;8.222030873995793
+2015/07/20 00:00:00 UTC;9.280467103184773
+2015/07/21 00:00:00 UTC;10.390691374091642
+2015/07/22 00:00:00 UTC;11.549928704232768
+2015/07/23 00:00:00 UTC;12.755281604001926
+2015/07/24 00:00:00 UTC;14.00373731888537
+2015/07/25 00:00:00 UTC;15.292175359779018
+2015/07/26 00:00:00 UTC;16.61737530258622
+2015/07/27 00:00:00 UTC;17.976024837600736
+2015/07/28 00:00:00 UTC;19.364728048556042
+2015/07/29 00:00:00 UTC;20.78001390064742
+2015/07/30 00:00:00 UTC;22.218344916311217
+2015/07/31 00:00:00 UTC;23.676126017076612
+2015/08/01 00:00:00 UTC;25.149713509389404
+2015/08/02 00:00:00 UTC;26.635424191948353
+2015/08/03 00:00:00 UTC;28.129544561790254
+2015/08/04 00:00:00 UTC;29.62834009611328
+2015/08/05 00:00:00 UTC;31.128064586639297
+2015/08/06 00:00:00 UTC;32.624969503183394
+2015/08/07 00:00:00 UTC;34.11531336302722
+2015/08/08 00:00:00 UTC;35.59537108267727
+2015/08/09 00:00:00 UTC;37.061443288633534
\ No newline at end of file

scikit-learn-.py

+
+# coding: utf-8
+
+# In[130]:
+
+get_ipython().magic(u'matplotlib inline')
+
+import csv
+import requests
+import ggplot
+import matplotlib
+matplotlib.use('agg')
+matplotlib.style.use('ggplot')
+import pandas as pd
+import matplotlib.pyplot as plt
+import numpy as np
+import statsmodels as sm
+from sklearn import linear_model
+from sklearn.gaussian_process import regression_models
+from StringIO import StringIO
+
+# need ssh tunnel to use this url
+url = 'http://localhost:2200/erp5/portal_skins/erp5_r_order_prediction_tutorial/SaleOrderModule_getArimaPredictionList'
+response = requests.get(url, auth=('zope', 'password'))
+csv_file = StringIO()
+csv_file.write(response.text)
+csv_file.seek(0)
+
+ts = pd.read_csv(csv_file, sep=';',
+                 parse_dates=[0],
+                 infer_datetime_format=True,
+                 squeeze=True,
+                 index_col='date')
+
+X = (ts.index - ts.index[0]).days.reshape(-1, 1)
+y = ts.values
+
+X_train = X[:]
+X_test = X[20:]
+
+y_train = y[:]
+y_test = y[20:]
+
+# Linear Regression
+lr = linear_model.LinearRegression(normalize=True)
+lr.fit(X_train, y_train)
+print lr.score(X_test, y_test)
+
+plt.scatter(X_test, y_test, color='black')
+plt.plot(X_test, lr.predict(X_test), color='blue', linewidth=3)
+
+# ElasticNet Regression
+
+mtl = linear_model.ElasticNet(alpha=0.1, normalize=True)
+mtl.fit(X_train, y_train)
+plt.plot(X_test, mtl.predict(X_test), color='green', linewidth=2)
+
+# SGD Regression
+
+sgd = linear_model.SGDRegressor(shuffle=False, eta0=0.25)
+sgd.fit(X_train, y_train)
+plt.plot(X_test, sgd.predict(X_test), color='red', linewidth=1)
+
+plt.xticks()
+plt.yticks()
+plt.show()
\ No newline at end of file

statsmodels.py

+
+# coding: utf-8
+
+# # Basic setup and data (with plot)
+
+# In[1]:
+
+get_ipython().magic(u'matplotlib inline')
+
+import csv
+import requests
+import ggplot
+import matplotlib
+matplotlib.use('agg')
+matplotlib.style.use('ggplot')
+import pandas as pd
+import matplotlib.pyplot as plt
+import statsmodels.api as sm
+import numpy as np
+from statsmodels.tsa.arima_process import ArmaProcess
+from StringIO import StringIO
+
+# need ssh tunnel to use this url
+url = 'http://zope:runescape@localhost:2200/erp5/portal_skins/erp5_r_order_prediction_tutorial/SaleOrderModule_getArimaPredictionList'
+response = requests.get(url, auth=('zope', 'runescape'))
+csv_file = StringIO()
+csv_file.write(response.text)
+csv_file.seek(0)
+
+ts = pd.read_csv(csv_file, sep=';',
+                 parse_dates=[0],
+                 infer_datetime_format=True,
+                 squeeze=True,
+                 index_col='date')
+ts.plot()
+
+
+# # Graphics to analyze auto correlation and partial auto correlation
+
+# In[2]:
+
+fig = plt.figure(figsize=(12, 8))
+ax1 = fig.add_subplot(211)
+fig = sm.graphics.tsa.plot_acf(ts, lags=30, ax=ax1)
+
+ax2 = fig.add_subplot(212)
+fig = sm.graphics.tsa.plot_pacf(ts, lags=30, ax=ax2)
+
+
+# Looks like partial correlation going to 0 after lag 3. So **ARMA(3,0)** might be  an adequated model for out data.
+
+# # Minimum order selection for ARMA
+
+# In[28]:
+
+answers = [sm.tsa.arma_order_select_ic(ts, ic=['aic'], trend=trend, fit_kw={'method': 'css'}) for trend in ['c', 'nc']]
+for answer in answers:
+    print answer['aic_min_order']
+
+
+# # Check if series is stationary
+
+# Using the Augmented Dicker-Fuller test we can check if out data is stationary. For this, we need to get an **adf statistic** value __smaller__ (more negative) than the critical value at 5% for regression **c**, **ct** and **nc**.
+
+# In[15]:
+
+def check_stationarity(series, regression='c'):
+    regressions = {
+        'c':   'constant only',
+        'ct':  'constant and trend',
+        'ctt': 'constant, linear and quadratic trend',
+        'nc':  'no constant, no trend'
+    }
+    result = sm.tsa.adfuller(series, regression=regression, autolag='t-stat', store=True)
+    store = result[-1]
+    if store.adfstat > result[2]['5%']:
+        print 'Time series is not stationary with %s. See data below:' % regressions[regression]
+        print '  - adf statistics: ', store.adfstat
+        print '  - p-value: ', result[1]
+        print '  - critical values: ', store.critvalues
+        return False
+    else:
+        print 'Time series is stationary with regression %s! :D\n' % regressions[regression]
+        return True
+
+stationarity_tests = [check_stationarity(ts, regression) for regression in ['c', 'ct', 'ctt', 'nc']]
+diff_ts = ts.diff()
+diff_ts.plot()
+    
+
+
+# # Fit the ARIMA model and plot the data + forecast
+
+# In[32]:
+
+arima = sm.tsa.ARIMA(ts, (3,0,0))
+arima_fit = arima.fit()
+
+arima_fit.resid.plot()
+
+fig, ax = plt.subplots()
+ax = ts.plot(ax=ax)
+fig = arima_fit.plot_predict(100, 200, dynamic=True, ax=ax)
+plt.show()
+

[Jérome Perrin]

I think you can just do pd.read_csv(StringIO(response.text)...

[Jérome Perrin]

( but this part will not be in the final tutorial, so it is really not important )

[Jérome Perrin]

Since we'll probably include this code in the final tutorial, can you please add a comment on why this is done ?

[Douglas]

@jerome yup.