

In [278]:

    
import pandas as pd
import numpy as np
from fbprophet import Prophet

실제함수 input



In [280]:

    
rawArrayDatas=[["2017-08-11", "2017-08-12", "2017-08-13", "2017-08-14", "2017-08-15","2017-08-16"],
               [20.0, 30.0, 40.0, 50.0, 60.0,20.0]]
processId=12
forecastDay=4



In [281]:

bug fix1

rawArrayDatas -> rawArrayDatas[0] rawArrayDatas이 이차원 배열이어서 len(rawArrayDatas)=2가 되고, len(rawArrayDatas[0])가 5가 된다.



In [281]:

    
mockForecast={}
rmse={}
forecast=[]
realForecast={}

trainSize=int(len(rawArrayDatas[0]) * 0.7)
testSize=len(rawArrayDatas[0])-trainSize

bug fix1

rawArrayDatas -> rawArrayDatas[0] rawArrayDatas이 이차원 배열이어서 len(rawArrayDatas)=2가 되고, len(rawArrayDatas[0])가 5가 된다.



In [282]:

    
print(trainSize)
print(testSize)

4
2

전체 data 전처리



In [283]:

    
ds = rawArrayDatas[0]
y = list(np.log(rawArrayDatas[1]))
sales = list(zip(ds, y))



In [284]:

    
rawArrayDatas



In [285]:

    
sales









    Out[285]:





[('2017-08-11', 2.9957322735539909),
 ('2017-08-12', 3.4011973816621555),
 ('2017-08-13', 3.6888794541139363),
 ('2017-08-14', 3.912023005428146),
 ('2017-08-15', 4.0943445622221004),
 ('2017-08-16', 2.9957322735539909)]



In [286]:

    
print(type(ds), type(y))









    



<class 'list'> <class 'list'>



In [287]:

    
preprocessedData= pd.DataFrame(data = sales, columns=['ds', 'y'])



In [288]:

    
preprocessedData



In [289]:

    
model = Prophet()
model.fit(preprocessedData)









    



Disabling yearly seasonality. Run prophet with yearly_seasonality=True to override this.
Disabling weekly seasonality. Run prophet with weekly_seasonality=True to override this.






    Out[289]:





<fbprophet.forecaster.Prophet at 0x1e24a26a5c0>



In [289]:

    
future = model.make_future_dataframe(periods=forecastDay)
forecast = future[-forecastDay:]



In [296]:

    
# Python
forecast = model.predict(future)
forecast[['ds', 'yhat']].tail()



In [296]:

    
forecast









    Out[296]:







  
    
      
      ds
      t
      trend
      seasonal_lower
      seasonal_upper
      trend_lower
      trend_upper
      yhat_lower
      yhat_upper
      seasonal
      yhat
    
  
  
    
      0
      2017-08-11
      0.0
      3.388195
      0.0
      0.0
      3.388195
      3.388195
      2.757076
      3.982821
      0.0
      3.388195
    
    
      1
      2017-08-12
      0.2
      3.439559
      0.0
      0.0
      3.439559
      3.439559
      2.805742
      4.107820
      0.0
      3.439559
    
    
      2
      2017-08-13
      0.4
      3.490922
      0.0
      0.0
      3.490922
      3.490922
      2.831384
      4.092323
      0.0
      3.490922
    
    
      3
      2017-08-14
      0.6
      3.542286
      0.0
      0.0
      3.542286
      3.542286
      2.943238
      4.220501
      0.0
      3.542286
    
    
      4
      2017-08-15
      0.8
      3.593649
      0.0
      0.0
      3.593649
      3.593649
      2.973599
      4.178181
      0.0
      3.593649
    
    
      5
      2017-08-16
      1.0
      3.645012
      0.0
      0.0
      3.645012
      3.645012
      3.053761
      4.263229
      0.0
      3.645012
    
    
      6
      2017-08-17
      1.2
      3.696375
      0.0
      0.0
      3.696374
      3.696375
      3.044536
      4.283623
      0.0
      3.696375
    
    
      7
      2017-08-18
      1.4
      3.747737
      0.0
      0.0
      3.747736
      3.747738
      3.137553
      4.346987
      0.0
      3.747737
    
    
      8
      2017-08-19
      1.6
      3.799100
      0.0
      0.0
      3.799098
      3.799102
      3.140557
      4.420334
      0.0
      3.799100
    
    
      9
      2017-08-20
      1.8
      3.850462
      0.0
      0.0
      3.850459
      3.850466
      3.232568
      4.507354
      0.0
      3.850462



In [297]:

    
forecast



In [298]:

    
model.plot(forecast)



In [299]:

    
forecastData= [np.exp(y) for y in forecast['yhat'][-forecastDay:]]



In [300]:



In [305]:

    
print(forecastData)









    



[40.300930634691582, 42.424972686557076, 44.66096139987782, 47.014796872064757]



In [306]:

    
data=rawArrayDatas[1]+forecastData



In [307]:

    
data









    Out[307]:





[20.0,
 30.0,
 40.0,
 50.0,
 60.0,
 20.0,
 40.300930634691582,
 42.424972686557076,
 44.66096139987782,
 47.014796872064757]



In [265]:









    



[20.0, 30.0, 40.0, 50.0, 60.0, 20.0, <generator object <genexpr> at 0x000001E24A2881A8>]



In [200]:

    
ans=np.log10(10)



In [173]:

    
ans









    Out[173]:





1.0



In [174]:

    
np.exp(ans)









    Out[174]:





2.7182818284590451



In [ ]:



In [ ]:



In [177]:

    
data= [np.exp(y) for y in forecast['yhat']]



In [178]:

    
print(data)









    



[29.612466176387688, 31.173202134528449, 32.816199704563701, 34.545794582779671, 36.366548303339172, 38.283230540225226, 40.300930634691582, 42.424972686557076, 44.66096139987782, 47.014796872064757]



In [165]:

    
date= [d.strftime('%Y-%m-%d') for d in forecast['ds']]



In [159]:

    
date









    Out[159]:





['2017-08-11',
 '2017-08-12',
 '2017-08-13',
 '2017-08-14',
 '2017-08-15',
 '2017-08-16',
 '2017-08-17',
 '2017-08-18',
 '2017-08-19',
 '2017-08-20']



In [129]:

    
dateStamp = list(forecast['ds'][-forecastDay:])
dateStamp
date = [p.strftime('%Y-%m-%d') for p in dateStamp]
date









    Out[129]:





['2017-08-16', '2017-08-17', '2017-08-18', '2017-08-19']



In [130]:

    
realForecast['Bayseian'] = Bayseian(preprocessedData=XY, forecastDay=forecastDay)[0]









    



---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-130-991702695106> in <module>()
----> 1 realForecast['Bayseian'] = Bayseian(preprocessedData=XY, forecastDay=forecastDay)[0]

NameError: name 'Bayseian' is not defined



In [99]:

    
realForecast









    Out[99]:





{'Bayseian':           ds
 5 2017-08-16
 6 2017-08-17
 7 2017-08-18
 8 2017-08-19}



In [100]:



In [42]:

    
XY = PrepareBayseian(rawArrayDatas)



In [43]:

    
Bayseian(preprocessedData=XY, forecastDay=forecastDay)[0]









    



Disabling yearly seasonality. Run prophet with yearly_seasonality=True to override this.
Disabling weekly seasonality. Run prophet with weekly_seasonality=True to override this.






    Out[43]:







  
    
      
      ds
    
  
  
    
      5
      2017-08-16
    
    
      6
      2017-08-17
    
    
      7
      2017-08-18
    
    
      8
      2017-08-19



In [33]:

    
LearningModuleRunner(rawArrayDatas, processId, forecastDay)









    



Disabling yearly seasonality. Run prophet with yearly_seasonality=True to override this.
Disabling weekly seasonality. Run prophet with weekly_seasonality=True to override this.



In [40]:

    
realForecast









    Out[40]:





{}



In [49]:

    
type(realForecast['Bayseian'])









    Out[49]:





pandas.core.frame.DataFrame



In [ ]:



In [ ]:



In [2]:

    
import sys
print(sys.version)









    



3.5.3 |Continuum Analytics, Inc.| (default, May 15 2017, 10:43:23) [MSC v.1900 64 bit (AMD64)]



In [ ]:



In [63]:

    
#data 준비위한 하드코딩
#rawArrayDatas
df0=pd.read_csv('./data/397_replace0with1.csv')
df0['y'] = np.log(df0['y'])
ds=df0['ds']
y=df0['y']



In [4]:

    
#processId
processId=1
#정식적인 data input과정
rawArrayDatas=[['2016-01-01','2016-01-02','2016-01-03','2016-01-04','2016-01-05'],[10,10,12,13,14]]
ds=rawArrayDatas[0]
y=rawArrayDatas[1]
sales=list(zip(ds,y))
day=5



In [150]:

    
sales









    Out[150]:





[('2016-01-01', 10),
 ('2016-01-02', 10),
 ('2016-01-03', 12),
 ('2016-01-04', 13),
 ('2016-01-05', 14)]



In [ ]:



In [143]:

    
rawArrayDatas[:][:2]









    Out[143]:





[['2016-01-01', '2016-01-02', '2016-01-03', '2016-01-04', '2016-01-05'],
 [10, 10, 12, 13, 14]]



In [144]:

    
rawArrayDatas=[['2016-01-01','2016-01-02','2016-01-03','2016-01-04','2016-01-05'],[10,10,12,13,14]]
ds=rawArrayDatas[0]
y=rawArrayDatas[1]



In [147]:

    
sales = list(zip(rawArrayDatas[0], rawArrayDatas[1]))



In [146]:

    
y









    Out[146]:





[10, 10, 12, 13, 14]



In [148]:

    
sales









    Out[148]:





[('2016-01-01', 10),
 ('2016-01-02', 10),
 ('2016-01-03', 12),
 ('2016-01-04', 13),
 ('2016-01-05', 14)]



In [124]:

    
ds=rawArrayDatas[0]
#-->year, month, dayOfWeek 추출

year=[2016,2016, 2016, 2017]
month=[1,1,1,1]
dayOfWeek=[1,2,3,4]
y=rawArrayDatas[1][:len(train)]



In [ ]:

    
ds=rawArrayDatas[0]
#-->year, month, dayOfWeek 추출

year=np.random.beta(2000, 2017, len(train))*(2017-2000)
month=np.random.beta(1, 12, len(train))*(12-1)
dayOfWeek=np.random.beta(0, 6, len(train))*(6-0)
y=rawArrayDatas[1][:len(train)]



In [152]:

    
year









    Out[152]:





[2016, 2016, 2016, 2017]



In [153]:

    
month









    Out[153]:





[1, 1, 1, 1]



In [154]:

    
dayOfWeek









    Out[154]:





[1, 2, 3, 4]



In [155]:

    
sales=list(zip(year, month, dayOfWeek, y))



In [156]:

    
sales









    Out[156]:





[(2016, 1, 1, 10), (2016, 1, 2, 10), (2016, 1, 3, 12), (2017, 1, 4, 13)]



In [157]:

    
x = pd.DataFrame(data = sales, columns=['year', 'month', 'dayOfWeek','y'])



In [158]:

    
x



In [125]:

    
y









    Out[125]:





[10, 10, 12, 13]



In [108]:

    
np.size(y)









    Out[108]:





392



In [126]:

    
x = pd.DataFrame(data = sales, columns=['year', 'month', 'dayOfWeek','y'])
x['month']









    Out[126]:





0    1
Name: month, dtype: int64



In [66]:

    
type(rawArrayDatas)









    Out[66]:





list



In [127]:

    
x









    Out[127]:







  
    
      
      year
      month
      dayOfWeek
      y
    
  
  
    
      0
      2016
      1
      1
      0      3.526361
1      3.713572
2      3.98898...



In [ ]:



In [62]:

    
type(x)









    Out[62]:





pandas.core.frame.DataFrame



In [49]:

    
list(df)









    Out[49]:





['ds', 'y']



In [ ]:



In [34]:

    
#값에 0이 있으면 log를 할 때 inf가 되므로 Initialization failed. 오류가 나니 주의할 것.
m = Prophet()
m.fit(df);









    



Disabling yearly seasonality. Run prophet with yearly_seasonality=True to override this.



In [53]:

    
future = m.make_future_dataframe(periods=day)
future.tail()



In [54]:

    
future[-day:]



In [159]:

    
future









    Out[159]:







  
    
      
      ds
    
  
  
    
      0
      2016-01-01
    
    
      1
      2016-01-02
    
    
      2
      2016-01-03
    
    
      3
      2016-01-04
    
    
      4
      2016-01-05
    
    
      5
      2016-01-06
    
    
      6
      2016-01-07
    
    
      7
      2016-01-08
    
    
      8
      2016-01-09
    
    
      9
      2016-01-10
    
    
      10
      2016-01-11
    
    
      11
      2016-01-12
    
    
      12
      2016-01-13
    
    
      13
      2016-01-14
    
    
      14
      2016-01-15
    
    
      15
      2016-01-16
    
    
      16
      2016-01-17
    
    
      17
      2016-01-18
    
    
      18
      2016-01-19
    
    
      19
      2016-01-20
    
    
      20
      2016-01-21
    
    
      21
      2016-01-22
    
    
      22
      2016-01-23
    
    
      23
      2016-01-24
    
    
      24
      2016-01-25
    
    
      25
      2016-01-26
    
    
      26
      2016-01-27
    
    
      27
      2016-01-28
    
    
      28
      2016-01-29
    
    
      29
      2016-01-30
    
    
      ...
      ...
    
    
      367
      2017-01-02
    
    
      368
      2017-01-03
    
    
      369
      2017-01-04
    
    
      370
      2017-01-05
    
    
      371
      2017-01-06
    
    
      372
      2017-01-07
    
    
      373
      2017-01-08
    
    
      374
      2017-01-09
    
    
      375
      2017-01-10
    
    
      376
      2017-01-11
    
    
      377
      2017-01-12
    
    
      378
      2017-01-13
    
    
      379
      2017-01-14
    
    
      380
      2017-01-15
    
    
      381
      2017-01-16
    
    
      382
      2017-01-17
    
    
      383
      2017-01-18
    
    
      384
      2017-01-19
    
    
      385
      2017-01-20
    
    
      386
      2017-01-21
    
    
      387
      2017-01-22
    
    
      388
      2017-01-23
    
    
      389
      2017-01-24
    
    
      390
      2017-01-25
    
    
      391
      2017-01-26
    
    
      392
      2017-01-27
    
    
      393
      2017-01-28
    
    
      394
      2017-01-29
    
    
      395
      2017-01-30
    
    
      396
      2017-01-31
    
  

397 rows × 1 columns



In [36]:

    
forecast=m.predict(future)
forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].tail()



In [166]:

    
temp=list(forecast['ds'][-day:])
date=[p.strftime('%Y-%m-%d') for p in temp]



In [169]:

    
date=[p.strftime('%Y-%m-%d') for p in temp]



In [170]:

    
date









    Out[170]:





['2017-01-27', '2017-01-28', '2017-01-29', '2017-01-30', '2017-01-31']



In [ ]:



In [ ]:



In [ ]:

    
strftime('We are the %d, %b %Y')



In [37]:

    
m.plot(forecast)









    Out[37]:



In [38]:

    
m.plot_components(forecast)









    Out[38]:



In [ ]:



In [ ]:



In [ ]:

	ds	yhat	yhat_lower	yhat_upper
392	2017-01-27	3.826318	3.098217	4.562641
393	2017-01-28	3.836739	3.115983	4.572362
394	2017-01-29	3.741370	3.053969	4.465159
395	2017-01-30	3.965737	3.280206	4.620421
396	2017-01-31	3.850588	3.135154	4.531074

	ds	y
0	2017-08-11	2.995732
1	2017-08-12	3.401197
2	2017-08-13	3.688879
3	2017-08-14	3.912023
4	2017-08-15	4.094345
5	2017-08-16	2.995732

	ds	yhat
5	2017-08-16	3.645012
6	2017-08-17	3.696375
7	2017-08-18	3.747737
8	2017-08-19	3.799100
9	2017-08-20	3.850462

	ds	t	trend	trend_lower	trend_upper	yhat_lower	yhat_upper	yhat
0	2017-08-11	0.0	3.388195	3.388195	3.388195	2.757076	3.982821	3.388195
1	2017-08-12	0.2	3.439559	3.439559	3.439559	2.805742	4.107820	3.439559
2	2017-08-13	0.4	3.490922	3.490922	3.490922	2.831384	4.092323	3.490922
3	2017-08-14	0.6	3.542286	3.542286	3.542286	2.943238	4.220501	3.542286
4	2017-08-15	0.8	3.593649	3.593649	3.593649	2.973599	4.178181	3.593649
5	2017-08-16	1.0	3.645012	3.645012	3.645012	3.053761	4.263229	3.645012
6	2017-08-17	1.2	3.696375	3.696374	3.696375	3.044536	4.283623	3.696375
7	2017-08-18	1.4	3.747737	3.747736	3.747738	3.137553	4.346987	3.747737
8	2017-08-19	1.6	3.799100	3.799098	3.799102	3.140557	4.420334	3.799100
9	2017-08-20	1.8	3.850462	3.850459	3.850466	3.232568	4.507354	3.850462

	ds
0	2016-01-01
1	2016-01-02
2	2016-01-03
3	2016-01-04
4	2016-01-05
5	2016-01-06
6	2016-01-07
7	2016-01-08
8	2016-01-09
9	2016-01-10
10	2016-01-11
11	2016-01-12
12	2016-01-13
13	2016-01-14
14	2016-01-15
15	2016-01-16
16	2016-01-17
17	2016-01-18
18	2016-01-19
19	2016-01-20
20	2016-01-21
21	2016-01-22
22	2016-01-23
23	2016-01-24
24	2016-01-25
25	2016-01-26
26	2016-01-27
27	2016-01-28
28	2016-01-29
29	2016-01-30
...	...
367	2017-01-02
368	2017-01-03
369	2017-01-04
370	2017-01-05
371	2017-01-06
372	2017-01-07
373	2017-01-08
374	2017-01-09
375	2017-01-10
376	2017-01-11
377	2017-01-12
378	2017-01-13
379	2017-01-14
380	2017-01-15
381	2017-01-16
382	2017-01-17
383	2017-01-18
384	2017-01-19
385	2017-01-20
386	2017-01-21
387	2017-01-22
388	2017-01-23
389	2017-01-24
390	2017-01-25
391	2017-01-26
392	2017-01-27
393	2017-01-28
394	2017-01-29
395	2017-01-30
396	2017-01-31