Chapter08.Times Series 쇼핑몰 매출 예측 (Times Series)

파이썬을 활용한 이커머스 데이터분석_강의를 듣고 따라했던 코딩과 요점을 정리하였다.

• 출처: fast campus

Chapter.08 쇼핑몰 매출 예측 (Times Series)¶

분석의 목적¶

• 시간 관련 데이터 다양하게 조작하기

• fbprophet을 통한 시계열 예측

• AutoRegressive 알고리즘을 통한 예측

1. 시간 관련 데이터 다양하게 조작하기¶

In [1]:

import pandas as pd
import numpy as np

In [2]:

from datetime import datetime, timedelta

In [3]:

today =datetime.now()

In [4]:

today

Out[4]:

datetime.datetime(2021, 7, 8, 8, 49, 6, 223234)

In [5]:

today.day

Out[5]:

8

In [6]:

today.year

Out[6]:

2021

In [7]:

today.month

Out[7]:

7

In [8]:

today

Out[8]:

datetime.datetime(2021, 7, 8, 8, 49, 6, 223234)

timedelta 의 파라미터값을 조정 할 수 있다. 년, 월, 일, 시간 ..

In [9]:

today + timedelta(days = 10)

Out[9]:

datetime.datetime(2021, 7, 18, 8, 49, 6, 223234)

간단하게 DataFrame 을 만들어 보자. (의미는 없는 값)

In [10]:

data = pd.DataFrame({'date':['7/6/2021', '7/7/2021', '7/8/2021'], 'value':[10,20,30]})

In [11]:

data

Out[11]:

	date	value
0	7/6/2021	10
1	7/7/2021	20
2	7/8/2021	30

In [12]:

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3 entries, 0 to 2
Data columns (total 2 columns):
 #   Column  Non-Null Count  Dtype 
---  ------  --------------  ----- 
 0   date    3 non-null      object
 1   value   3 non-null      int64 
dtypes: int64(1), object(1)
memory usage: 176.0+ bytes

Dtype을 보면 date가 object로 되어있다. string(문자열)로 인식 하고 있기 때문에 작업하는데에 에로사항이 많기에 바꾸어 주도록 하자.

• pd.to_datetime() 함수를 사용한다.

In [13]:

pd.to_datetime(data['date'])

Out[13]:

0   2021-07-06
1   2021-07-07
2   2021-07-08
Name: date, dtype: datetime64[ns]

dtype 가 object에서 datetime64로 변경 되었다.

In [14]:

data['date'] = pd.to_datetime(data['date'])

In [15]:

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3 entries, 0 to 2
Data columns (total 2 columns):
 #   Column  Non-Null Count  Dtype         
---  ------  --------------  -----         
 0   date    3 non-null      datetime64[ns]
 1   value   3 non-null      int64         
dtypes: datetime64[ns](1), int64(1)
memory usage: 176.0 bytes

In [16]:

data['date'].loc[0]

Out[16]:

Timestamp('2021-07-06 00:00:00')

In [17]:

data['date'].loc[0].month

Out[17]:

7

In [18]:

data = pd.DataFrame({'date':['7-6-2021 1:11:11', '7-7-2021 2:11:11', '7-8-2021 3:11:11'], 'value':[10,20,30]})

In [19]:

data

Out[19]:

	date	value
0	7-6-2021 1:11:11	10
1	7-7-2021 2:11:11	20
2	7-8-2021 3:11:11	30

In [20]:

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3 entries, 0 to 2
Data columns (total 2 columns):
 #   Column  Non-Null Count  Dtype 
---  ------  --------------  ----- 
 0   date    3 non-null      object
 1   value   3 non-null      int64 
dtypes: int64(1), object(1)
memory usage: 176.0+ bytes

In [21]:

pd.to_datetime(data['date'])

Out[21]:

0   2021-07-06 01:11:11
1   2021-07-07 02:11:11
2   2021-07-08 03:11:11
Name: date, dtype: datetime64[ns]

위와 했던 방법과 동일하게 바꾸어 주었다. dtype 변경 그리고 년-월-일 순으로 똑똑하게 알아서 바꿔준다는것을 눈으로 확인 해 볼 수 있다.

In [22]:

data['date'] = pd.to_datetime(data['date'])

In [23]:

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3 entries, 0 to 2
Data columns (total 2 columns):
 #   Column  Non-Null Count  Dtype         
---  ------  --------------  -----         
 0   date    3 non-null      datetime64[ns]
 1   value   3 non-null      int64         
dtypes: datetime64[ns](1), int64(1)
memory usage: 176.0 bytes

In [24]:

data['date'].loc[0]

Out[24]:

Timestamp('2021-07-06 01:11:11')

In [25]:

data['date'].loc[0].day

Out[25]:

6

In [26]:

data['date'].loc[0].second

Out[26]:

11

In [27]:

data = pd.DataFrame({'date': ['21-7-8', '21-7-9', '21-7-10'], 'value': [10,20,30] })

In [28]:

data

Out[28]:

	date	value
0	21-7-8	10
1	21-7-9	20
2	21-7-10	30

In [29]:

pd.to_datetime(data['date'])

Out[29]:

0   2008-07-21
1   2009-07-21
2   2010-07-21
Name: date, dtype: datetime64[ns]

미국식 날짜표현법 으로 나타났다. format 파라미터를 추가하여 이것을 재수정 해보자.

In [30]:

pd.to_datetime(data['date'], format = '%y-%m-%d')

Out[30]:

0   2021-07-08
1   2021-07-09
2   2021-07-10
Name: date, dtype: datetime64[ns]

이렇게 우리가 원하는 형식으로 날짜표현법을 바꿀 수 있다.

마지막으로 datetime 컬럼에 관하여 알아볼 것이 있다. 엑셀 파일을 불러와 보자.

In [31]:

data = pd.read_excel('./data/Superstore.xls')

In [32]:

data.head()

Out[32]:

	Row ID	Order ID	Order Date	Ship Date	Ship Mode	Customer ID	Customer Name	Segment	Country	City	...	Postal Code	Region	Product ID	Category	Sub-Category	Product Name	Sales	Quantity	Discount	Profit
0	1	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	...	42420	South	FUR-BO-10001798	Furniture	Bookcases	Bush Somerset Collection Bookcase	261.9600	2	0.00	41.9136
1	2	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	...	42420	South	FUR-CH-10000454	Furniture	Chairs	Hon Deluxe Fabric Upholstered Stacking Chairs,...	731.9400	3	0.00	219.5820
2	3	CA-2016-138688	2016-06-12	2016-06-16	Second Class	DV-13045	Darrin Van Huff	Corporate	United States	Los Angeles	...	90036	West	OFF-LA-10000240	Office Supplies	Labels	Self-Adhesive Address Labels for Typewriters b...	14.6200	2	0.00	6.8714
3	4	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	...	33311	South	FUR-TA-10000577	Furniture	Tables	Bretford CR4500 Series Slim Rectangular Table	957.5775	5	0.45	-383.0310
4	5	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	...	33311	South	OFF-ST-10000760	Office Supplies	Storage	Eldon Fold 'N Roll Cart System	22.3680	2	0.20	2.5164

5 rows × 21 columns

data가 깔끔한 컬럼만 인덱싱해 보자.

In [33]:

data = data[['Order Date', 'Ship Date', 'Sales', 'Profit']]

In [34]:

data

Out[34]:

	Order Date	Ship Date	Sales	Profit
0	2016-11-08	2016-11-11	261.9600	41.9136
1	2016-11-08	2016-11-11	731.9400	219.5820
2	2016-06-12	2016-06-16	14.6200	6.8714
3	2015-10-11	2015-10-18	957.5775	-383.0310
4	2015-10-11	2015-10-18	22.3680	2.5164
...	...	...	...	...
9989	2014-01-21	2014-01-23	25.2480	4.1028
9990	2017-02-26	2017-03-03	91.9600	15.6332
9991	2017-02-26	2017-03-03	258.5760	19.3932
9992	2017-02-26	2017-03-03	29.6000	13.3200
9993	2017-05-04	2017-05-09	243.1600	72.9480

9994 rows × 4 columns

In [35]:

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 9994 entries, 0 to 9993
Data columns (total 4 columns):
 #   Column      Non-Null Count  Dtype         
---  ------      --------------  -----         
 0   Order Date  9994 non-null   datetime64[ns]
 1   Ship Date   9994 non-null   datetime64[ns]
 2   Sales       9994 non-null   float64       
 3   Profit      9994 non-null   float64       
dtypes: datetime64[ns](2), float64(2)
memory usage: 312.4 KB

여기서 확인해야 할 부분은 Date 컬럼의 Dtype가 datetime으로 나오는지. string으로 불러오질 경우는 pd.to_datetime() 함수를 사용한다.

같은 datetime은 '-' 이 될까? Order Date 와 Ship Date 를 빼보자. (주문 날짜와 배송 날짜)

In [36]:

data['gap'] = data['Ship Date'] - data['Order Date']

In [37]:

data

Out[37]:

	Order Date	Ship Date	Sales	Profit	gap
0	2016-11-08	2016-11-11	261.9600	41.9136	3 days
1	2016-11-08	2016-11-11	731.9400	219.5820	3 days
2	2016-06-12	2016-06-16	14.6200	6.8714	4 days
3	2015-10-11	2015-10-18	957.5775	-383.0310	7 days
4	2015-10-11	2015-10-18	22.3680	2.5164	7 days
...	...	...	...	...	...
9989	2014-01-21	2014-01-23	25.2480	4.1028	2 days
9990	2017-02-26	2017-03-03	91.9600	15.6332	5 days
9991	2017-02-26	2017-03-03	258.5760	19.3932	5 days
9992	2017-02-26	2017-03-03	29.6000	13.3200	5 days
9993	2017-05-04	2017-05-09	243.1600	72.9480	5 days

9994 rows × 5 columns

Linear Regression을 한번 만들어 보자. (test용)

In [38]:

from sklearn.linear_model import LinearRegression

In [39]:

model = LinearRegression()

In [40]:

X = data[['Profit', 'gap']]
y = data['Sales']

In [41]:

X

Out[41]:

	Profit	gap
0	41.9136	3 days
1	219.5820	3 days
2	6.8714	4 days
3	-383.0310	7 days
4	2.5164	7 days
...	...	...
9989	4.1028	2 days
9990	15.6332	5 days
9991	19.3932	5 days
9992	13.3200	5 days
9993	72.9480	5 days

9994 rows × 2 columns

In [42]:

X.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 9994 entries, 0 to 9993
Data columns (total 2 columns):
 #   Column  Non-Null Count  Dtype          
---  ------  --------------  -----          
 0   Profit  9994 non-null   float64        
 1   gap     9994 non-null   timedelta64[ns]
dtypes: float64(1), timedelta64[ns](1)
memory usage: 156.3 KB

gap 컬럼은 timedelta64[ns] 으로 나타나는데, 독립변수로 쓸 수 있을지 확인해 보는 과정이다.

In [43]:

model.fit(X, y)

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-43-6d10fe8973eb> in <module>
----> 1 model.fit(X, y)

~/.local/lib/python3.6/site-packages/sklearn/linear_model/_base.py in fit(self, X, y, sample_weight)
    517 
    518         X, y = self._validate_data(X, y, accept_sparse=accept_sparse,
--> 519                                    y_numeric=True, multi_output=True)
    520 
    521         if sample_weight is not None:

~/.local/lib/python3.6/site-packages/sklearn/base.py in _validate_data(self, X, y, reset, validate_separately, **check_params)
    431                 y = check_array(y, **check_y_params)
    432             else:
--> 433                 X, y = check_X_y(X, y, **check_params)
    434             out = X, y
    435 

~/.local/lib/python3.6/site-packages/sklearn/utils/validation.py in inner_f(*args, **kwargs)
     61             extra_args = len(args) - len(all_args)
     62             if extra_args <= 0:
---> 63                 return f(*args, **kwargs)
     64 
     65             # extra_args > 0

~/.local/lib/python3.6/site-packages/sklearn/utils/validation.py in check_X_y(X, y, accept_sparse, accept_large_sparse, dtype, order, copy, force_all_finite, ensure_2d, allow_nd, multi_output, ensure_min_samples, ensure_min_features, y_numeric, estimator)
    876                     ensure_min_samples=ensure_min_samples,
    877                     ensure_min_features=ensure_min_features,
--> 878                     estimator=estimator)
    879     if multi_output:
    880         y = check_array(y, accept_sparse='csr', force_all_finite=True,

~/.local/lib/python3.6/site-packages/sklearn/utils/validation.py in inner_f(*args, **kwargs)
     61             extra_args = len(args) - len(all_args)
     62             if extra_args <= 0:
---> 63                 return f(*args, **kwargs)
     64 
     65             # extra_args > 0

~/.local/lib/python3.6/site-packages/sklearn/utils/validation.py in check_array(array, accept_sparse, accept_large_sparse, dtype, order, copy, force_all_finite, ensure_2d, allow_nd, ensure_min_samples, ensure_min_features, estimator)
    595 
    596         if all(isinstance(dtype, np.dtype) for dtype in dtypes_orig):
--> 597             dtype_orig = np.result_type(*dtypes_orig)
    598 
    599     if dtype_numeric:

<__array_function__ internals> in result_type(*args, **kwargs)

TypeError: invalid type promotion

gap 컬럼은 timedelta64[ns] 으로 나타나는데, 독립변수로 쓸 수 없다. 에러가 나타났다. 이를 해결해 보자.

• 첫번째 방법 : .apply(lambda) 사용

In [44]:

X['gap'] = X['gap'].apply(lambda x: x.days)

/home/ubuntu/.local/lib/python3.6/site-packages/ipykernel_launcher.py:1: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  """Entry point for launching an IPython kernel.

In [45]:

X

Out[45]:

	Profit	gap
0	41.9136	3
1	219.5820	3
2	6.8714	4
3	-383.0310	7
4	2.5164	7
...	...	...
9989	4.1028	2
9990	15.6332	5
9991	19.3932	5
9992	13.3200	5
9993	72.9480	5

9994 rows × 2 columns

In [46]:

model.fit(X, y)

Out[46]:

LinearRegression()

In [47]:

model.predict(X)

Out[47]:

array([248.5051552 , 474.93985046, 202.01637137, ..., 216.14692021,
       208.40675263, 284.40140338])

• 두번째 방법 : datetime 형태를 숫자형태로 바꿔주는 방법 -> 모든 시간을 '초' 로 나타내는 방법 / Unix Time stamp

In [48]:

data = pd.read_excel('./data/Superstore.xls')

In [49]:

data.head()

Out[49]:

	Row ID	Order ID	Order Date	Ship Date	Ship Mode	Customer ID	Customer Name	Segment	Country	City	...	Postal Code	Region	Product ID	Category	Sub-Category	Product Name	Sales	Quantity	Discount	Profit
0	1	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	...	42420	South	FUR-BO-10001798	Furniture	Bookcases	Bush Somerset Collection Bookcase	261.9600	2	0.00	41.9136
1	2	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	...	42420	South	FUR-CH-10000454	Furniture	Chairs	Hon Deluxe Fabric Upholstered Stacking Chairs,...	731.9400	3	0.00	219.5820
2	3	CA-2016-138688	2016-06-12	2016-06-16	Second Class	DV-13045	Darrin Van Huff	Corporate	United States	Los Angeles	...	90036	West	OFF-LA-10000240	Office Supplies	Labels	Self-Adhesive Address Labels for Typewriters b...	14.6200	2	0.00	6.8714
3	4	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	...	33311	South	FUR-TA-10000577	Furniture	Tables	Bretford CR4500 Series Slim Rectangular Table	957.5775	5	0.45	-383.0310
4	5	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	...	33311	South	OFF-ST-10000760	Office Supplies	Storage	Eldon Fold 'N Roll Cart System	22.3680	2	0.20	2.5164

5 rows × 21 columns

data가 깔끔한 컬럼만 인덱싱해 보자.

In [50]:

data = data[['Order Date', 'Ship Date', 'Sales', 'Profit']]

In [51]:

data

Out[51]:

	Order Date	Ship Date	Sales	Profit
0	2016-11-08	2016-11-11	261.9600	41.9136
1	2016-11-08	2016-11-11	731.9400	219.5820
2	2016-06-12	2016-06-16	14.6200	6.8714
3	2015-10-11	2015-10-18	957.5775	-383.0310
4	2015-10-11	2015-10-18	22.3680	2.5164
...	...	...	...	...
9989	2014-01-21	2014-01-23	25.2480	4.1028
9990	2017-02-26	2017-03-03	91.9600	15.6332
9991	2017-02-26	2017-03-03	258.5760	19.3932
9992	2017-02-26	2017-03-03	29.6000	13.3200
9993	2017-05-04	2017-05-09	243.1600	72.9480

9994 rows × 4 columns

In [52]:

pd.to_numeric(data['Order Date'])

Out[52]:

0       1478563200000000000
1       1478563200000000000
2       1465689600000000000
3       1444521600000000000
4       1444521600000000000
               ...         
9989    1390262400000000000
9990    1488067200000000000
9991    1488067200000000000
9992    1488067200000000000
9993    1493856000000000000
Name: Order Date, Length: 9994, dtype: int64

0이 너무 많다. 제거해주자. 불필요하기때문에

In [53]:

pd.to_numeric(data['Order Date']) / 10**9

Out[53]:

0       1.478563e+09
1       1.478563e+09
2       1.465690e+09
3       1.444522e+09
4       1.444522e+09
            ...     
9989    1.390262e+09
9990    1.488067e+09
9991    1.488067e+09
9992    1.488067e+09
9993    1.493856e+09
Name: Order Date, Length: 9994, dtype: float64

In [54]:

pd.set_option('display.float_format', lambda x: '%.0f' % x)

In [55]:

pd.to_numeric(data['Order Date']) / 10**9

Out[55]:

0      1478563200
1      1478563200
2      1465689600
3      1444521600
4      1444521600
          ...    
9989   1390262400
9990   1488067200
9991   1488067200
9992   1488067200
9993   1493856000
Name: Order Date, Length: 9994, dtype: float64

In [56]:

data['Order Date'] =  pd.to_numeric(data['Order Date']) / 10**9
data['Ship Date'] = pd.to_numeric(data['Ship Date']) / 10**9

In [57]:

data

Out[57]:

	Order Date	Ship Date	Sales	Profit
0	1478563200	1478822400	262	42
1	1478563200	1478822400	732	220
2	1465689600	1466035200	15	7
3	1444521600	1445126400	958	-383
4	1444521600	1445126400	22	3
...	...	...	...	...
9989	1390262400	1390435200	25	4
9990	1488067200	1488499200	92	16
9991	1488067200	1488499200	259	19
9992	1488067200	1488499200	30	13
9993	1493856000	1494288000	243	73

9994 rows × 4 columns

In [60]:

data['gap'] = data['Ship Date'] - data['Order Date']

In [61]:

data

Out[61]:

	Order Date	Ship Date	Sales	Profit	gap
0	1478563200	1478822400	262	42	259200
1	1478563200	1478822400	732	220	259200
2	1465689600	1466035200	15	7	345600
3	1444521600	1445126400	958	-383	604800
4	1444521600	1445126400	22	3	604800
...	...	...	...	...	...
9989	1390262400	1390435200	25	4	172800
9990	1488067200	1488499200	92	16	432000
9991	1488067200	1488499200	259	19	432000
9992	1488067200	1488499200	30	13	432000
9993	1493856000	1494288000	243	73	432000

9994 rows × 5 columns

In [62]:

X = data[['Profit', 'gap']]
y = data['Sales']

In [63]:

model = LinearRegression()
model.fit(X,y)

Out[63]:

LinearRegression()

에러가 나타나지 않고 잘 돌아가는 것을 확인 할 수 있다. 여기까지 1.시간데이터 다양하게 조작하기 를 알아보았다.

2. fbprophet을 통한 시계열 예측¶

In [64]:

!pip install pystan

Collecting pystan
  Downloading https://files.pythonhosted.org/packages/f3/67/983a2de85fe82b7c2813fe01801edca19c6f193db8cc5903e684f77ce870/pystan-2.19.1.1-cp27-cp27mu-manylinux1_x86_64.whl (67.2MB)
    100% |████████████████████████████████| 67.2MB 17kB/s  eta 0:00:01
Collecting numpy>=1.7 (from pystan)
  Downloading https://files.pythonhosted.org/packages/3a/5f/47e578b3ae79e2624e205445ab77a1848acdaa2929a00eeef6b16eaaeb20/numpy-1.16.6-cp27-cp27mu-manylinux1_x86_64.whl (17.0MB)
    100% |████████████████████████████████| 17.0MB 70kB/s  eta 0:00:01
Collecting Cython!=0.25.1,>=0.22 (from pystan)
  Downloading https://files.pythonhosted.org/packages/ae/85/9a00bfcd16d481676f2c78550c3f5352197c0436ae6fb9af445242be0edc/Cython-0.29.23-cp27-cp27mu-manylinux1_x86_64.whl (1.9MB)
    100% |████████████████████████████████| 1.9MB 648kB/s eta 0:00:01
Installing collected packages: numpy, Cython, pystan
Successfully installed Cython-0.29.23 numpy-1.16.6 pystan-2.19.1.1

In [65]:

!pip install plotly

Collecting plotly
  Using cached https://files.pythonhosted.org/packages/1f/f6/bd3c17c8003b6641df1228e80e1acac97ed8402635e46c2571f8e1ef63af/plotly-4.14.3-py2.py3-none-any.whl
Collecting six (from plotly)
  Using cached https://files.pythonhosted.org/packages/d9/5a/e7c31adbe875f2abbb91bd84cf2dc52d792b5a01506781dbcf25c91daf11/six-1.16.0-py2.py3-none-any.whl
Collecting retrying>=1.3.3 (from plotly)
  Using cached https://files.pythonhosted.org/packages/44/ef/beae4b4ef80902f22e3af073397f079c96969c69b2c7d52a57ea9ae61c9d/retrying-1.3.3.tar.gz
Building wheels for collected packages: retrying
  Running setup.py bdist_wheel for retrying ... done
  Stored in directory: /home/ubuntu/.cache/pip/wheels/d7/a9/33/acc7b709e2a35caa7d4cae442f6fe6fbf2c43f80823d46460c
Successfully built retrying
Installing collected packages: six, retrying, plotly
Successfully installed plotly-4.14.3 retrying-1.3.3 six-1.16.0

In [1]:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

In [6]:

data = pd.read_excel('./data/Superstore.xls', index_col=0)

In [7]:

data.head()

Out[7]:

	Order ID	Order Date	Ship Date	Ship Mode	Customer ID	Customer Name	Segment	Country	City	State	Postal Code	Region	Product ID	Category	Sub-Category	Product Name	Sales	Quantity	Discount	Profit
Row ID
1	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-BO-10001798	Furniture	Bookcases	Bush Somerset Collection Bookcase	261.9600	2	0.00	41.9136
2	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-CH-10000454	Furniture	Chairs	Hon Deluxe Fabric Upholstered Stacking Chairs,...	731.9400	3	0.00	219.5820
3	CA-2016-138688	2016-06-12	2016-06-16	Second Class	DV-13045	Darrin Van Huff	Corporate	United States	Los Angeles	California	90036	West	OFF-LA-10000240	Office Supplies	Labels	Self-Adhesive Address Labels for Typewriters b...	14.6200	2	0.00	6.8714
4	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	FUR-TA-10000577	Furniture	Tables	Bretford CR4500 Series Slim Rectangular Table	957.5775	5	0.45	-383.0310
5	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	OFF-ST-10000760	Office Supplies	Storage	Eldon Fold 'N Roll Cart System	22.3680	2	0.20	2.5164

Times Series의 경우 컬럼을 많이 살펴 볼 필요는 없다. 우리가 보고자하는 컬럼들만 보자.

In [8]:

data.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 9994 entries, 1 to 9994
Data columns (total 20 columns):
 #   Column         Non-Null Count  Dtype         
---  ------         --------------  -----         
 0   Order ID       9994 non-null   object        
 1   Order Date     9994 non-null   datetime64[ns]
 2   Ship Date      9994 non-null   datetime64[ns]
 3   Ship Mode      9994 non-null   object        
 4   Customer ID    9994 non-null   object        
 5   Customer Name  9994 non-null   object        
 6   Segment        9994 non-null   object        
 7   Country        9994 non-null   object        
 8   City           9994 non-null   object        
 9   State          9994 non-null   object        
 10  Postal Code    9994 non-null   int64         
 11  Region         9994 non-null   object        
 12  Product ID     9994 non-null   object        
 13  Category       9994 non-null   object        
 14  Sub-Category   9994 non-null   object        
 15  Product Name   9994 non-null   object        
 16  Sales          9994 non-null   float64       
 17  Quantity       9994 non-null   int64         
 18  Discount       9994 non-null   float64       
 19  Profit         9994 non-null   float64       
dtypes: datetime64[ns](2), float64(3), int64(2), object(13)
memory usage: 1.6+ MB

결측치는 없고, Dtype를 확인한다.

In [9]:

data.describe()

Out[9]:

	Postal Code	Sales	Quantity	Discount	Profit
count	9994.000000	9994.000000	9994.000000	9994.000000	9994.000000
mean	55190.379428	229.858001	3.789574	0.156203	28.656896
std	32063.693350	623.245101	2.225110	0.206452	234.260108
min	1040.000000	0.444000	1.000000	0.000000	-6599.978000
25%	23223.000000	17.280000	2.000000	0.000000	1.728750
50%	56430.500000	54.490000	3.000000	0.200000	8.666500
75%	90008.000000	209.940000	5.000000	0.200000	29.364000
max	99301.000000	22638.480000	14.000000	0.800000	8399.976000

Sales에서 max 값은 아웃라이너 같다고 예상해볼 수 있다. 너무 값이 크기때문에.

In [10]:

data['Category'].unique()

Out[10]:

array(['Furniture', 'Office Supplies', 'Technology'], dtype=object)

.unique() 함수를 사용하여 Category의 종류를 확인해 볼 수 있다.

In [11]:

data['Category'].value_counts()

Out[11]:

Office Supplies    6026
Furniture          2121
Technology         1847
Name: Category, dtype: int64

.value_counts() 함수를 사용하여 Category의 종류와 수량을 확인 할 수 있다.

In [14]:

data['Category'] == 'Furniture'  ## 불리안 True or False 확인

Out[14]:

Row ID
1        True
2        True
3       False
4        True
5       False
        ...  
9990     True
9991     True
9992    False
9993    False
9994    False
Name: Category, Length: 9994, dtype: bool

In [16]:

data.head() # 위 코드와 대조해서 확인 해본다. 

Out[16]:

	Order ID	Order Date	Ship Date	Ship Mode	Customer ID	Customer Name	Segment	Country	City	State	Postal Code	Region	Product ID	Category	Sub-Category	Product Name	Sales	Quantity	Discount	Profit
Row ID
1	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-BO-10001798	Furniture	Bookcases	Bush Somerset Collection Bookcase	261.9600	2	0.00	41.9136
2	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-CH-10000454	Furniture	Chairs	Hon Deluxe Fabric Upholstered Stacking Chairs,...	731.9400	3	0.00	219.5820
3	CA-2016-138688	2016-06-12	2016-06-16	Second Class	DV-13045	Darrin Van Huff	Corporate	United States	Los Angeles	California	90036	West	OFF-LA-10000240	Office Supplies	Labels	Self-Adhesive Address Labels for Typewriters b...	14.6200	2	0.00	6.8714
4	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	FUR-TA-10000577	Furniture	Tables	Bretford CR4500 Series Slim Rectangular Table	957.5775	5	0.45	-383.0310
5	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	OFF-ST-10000760	Office Supplies	Storage	Eldon Fold 'N Roll Cart System	22.3680	2	0.20	2.5164

In [18]:

data[data['Category'] == 'Furniture' ] # data['Category'] == 'Furniture' 을 data[]에 넣어 주면 True만 불러와주는 기능을 한다.

Out[18]:

	Order ID	Order Date	Ship Date	Ship Mode	Customer ID	Customer Name	Segment	Country	City	State	Postal Code	Region	Product ID	Category	Sub-Category	Product Name	Sales	Quantity	Discount	Profit
Row ID
1	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-BO-10001798	Furniture	Bookcases	Bush Somerset Collection Bookcase	261.9600	2	0.00	41.9136
2	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-CH-10000454	Furniture	Chairs	Hon Deluxe Fabric Upholstered Stacking Chairs,...	731.9400	3	0.00	219.5820
4	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	FUR-TA-10000577	Furniture	Tables	Bretford CR4500 Series Slim Rectangular Table	957.5775	5	0.45	-383.0310
6	CA-2014-115812	2014-06-09	2014-06-14	Standard Class	BH-11710	Brosina Hoffman	Consumer	United States	Los Angeles	California	90032	West	FUR-FU-10001487	Furniture	Furnishings	Eldon Expressions Wood and Plastic Desk Access...	48.8600	7	0.00	14.1694
11	CA-2014-115812	2014-06-09	2014-06-14	Standard Class	BH-11710	Brosina Hoffman	Consumer	United States	Los Angeles	California	90032	West	FUR-TA-10001539	Furniture	Tables	Chromcraft Rectangular Conference Tables	1706.1840	9	0.20	85.3092
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
9963	CA-2015-168088	2015-03-19	2015-03-22	First Class	CM-12655	Corinna Mitchell	Home Office	United States	Houston	Texas	77041	Central	FUR-BO-10004218	Furniture	Bookcases	Bush Heritage Pine Collection 5-Shelf Bookcase...	383.4656	4	0.32	-67.6704
9965	CA-2016-146374	2016-12-05	2016-12-10	Second Class	HE-14800	Harold Engle	Corporate	United States	Newark	Delaware	19711	East	FUR-FU-10002671	Furniture	Furnishings	Electrix 20W Halogen Replacement Bulb for Zoom...	13.4000	1	0.00	6.4320
9981	US-2015-151435	2015-09-06	2015-09-09	Second Class	SW-20455	Shaun Weien	Consumer	United States	Lafayette	Louisiana	70506	South	FUR-TA-10001039	Furniture	Tables	KI Adjustable-Height Table	85.9800	1	0.00	22.3548
9990	CA-2014-110422	2014-01-21	2014-01-23	Second Class	TB-21400	Tom Boeckenhauer	Consumer	United States	Miami	Florida	33180	South	FUR-FU-10001889	Furniture	Furnishings	Ultra Door Pull Handle	25.2480	3	0.20	4.1028
9991	CA-2017-121258	2017-02-26	2017-03-03	Standard Class	DB-13060	Dave Brooks	Consumer	United States	Costa Mesa	California	92627	West	FUR-FU-10000747	Furniture	Furnishings	Tenex B1-RE Series Chair Mats for Low Pile Car...	91.9600	2	0.00	15.6332

2121 rows × 20 columns

In [21]:

data = data[data['Category'] == 'Furniture']  # data에 할당

In [22]:

data.head()

Out[22]:

	Order ID	Order Date	Ship Date	Ship Mode	Customer ID	Customer Name	Segment	Country	City	State	Postal Code	Region	Product ID	Category	Sub-Category	Product Name	Sales	Quantity	Discount	Profit
Row ID
1	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-BO-10001798	Furniture	Bookcases	Bush Somerset Collection Bookcase	261.9600	2	0.00	41.9136
2	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-CH-10000454	Furniture	Chairs	Hon Deluxe Fabric Upholstered Stacking Chairs,...	731.9400	3	0.00	219.5820
4	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	FUR-TA-10000577	Furniture	Tables	Bretford CR4500 Series Slim Rectangular Table	957.5775	5	0.45	-383.0310
6	CA-2014-115812	2014-06-09	2014-06-14	Standard Class	BH-11710	Brosina Hoffman	Consumer	United States	Los Angeles	California	90032	West	FUR-FU-10001487	Furniture	Furnishings	Eldon Expressions Wood and Plastic Desk Access...	48.8600	7	0.00	14.1694
11	CA-2014-115812	2014-06-09	2014-06-14	Standard Class	BH-11710	Brosina Hoffman	Consumer	United States	Los Angeles	California	90032	West	FUR-TA-10001539	Furniture	Tables	Chromcraft Rectangular Conference Tables	1706.1840	9	0.20	85.3092

In [24]:

data = data[['Order Date', 'Sales']]

In [25]:

data

Out[25]:

	Order Date	Sales
Row ID
1	2016-11-08	261.9600
2	2016-11-08	731.9400
4	2015-10-11	957.5775
6	2014-06-09	48.8600
11	2014-06-09	1706.1840
...	...	...
9963	2015-03-19	383.4656
9965	2016-12-05	13.4000
9981	2015-09-06	85.9800
9990	2014-01-21	25.2480
9991	2017-02-26	91.9600

2121 rows × 2 columns

In [27]:

data.sort_values('Order Date')  #Order Date 칼럼 기준으로 오름차순으로 불러오는 함수를 사용

Out[27]:

	Order Date	Sales
Row ID
7475	2014-01-06	2573.820
7661	2014-01-07	76.728
867	2014-01-10	51.940
717	2014-01-11	9.940
2979	2014-01-13	545.940
...	...	...
7637	2017-12-29	258.750
7633	2017-12-29	1207.840
5132	2017-12-29	101.120
7636	2017-12-29	300.980
907	2017-12-30	323.136

2121 rows × 2 columns

우리가 알고자하는 것은 시간에 따른 Sales의 트렌드 임으로 다시 정리를 해준다.

In [31]:

data = data.groupby('Order Date').sum()  # 각 날짜별로 총합을 구한다.

In [30]:

data

Out[30]:

	Sales
Order Date
2014-01-06	2573.8200
2014-01-07	76.7280
2014-01-10	51.9400
2014-01-11	9.9400
2014-01-13	879.9390
...	...
2017-12-24	1393.4940
2017-12-25	832.4540
2017-12-28	551.2568
2017-12-29	2330.7180
2017-12-30	323.1360

889 rows × 1 columns

In [33]:

data.index

Out[33]:

DatetimeIndex(['2014-01-06', '2014-01-07', '2014-01-10', '2014-01-11',
               '2014-01-13', '2014-01-14', '2014-01-16', '2014-01-19',
               '2014-01-20', '2014-01-21',
               ...
               '2017-12-18', '2017-12-19', '2017-12-21', '2017-12-22',
               '2017-12-23', '2017-12-24', '2017-12-25', '2017-12-28',
               '2017-12-29', '2017-12-30'],
              dtype='datetime64[ns]', name='Order Date', length=889, freq=None)

plot을 그려보자.

In [34]:

plt.figure(figsize = (20,10))
sns.lineplot(x = data.index, y = data['Sales'])

Out[34]:

<AxesSubplot:xlabel='Order Date', ylabel='Sales'>

위와 같이 컬럼이 한가지만 존재할 경우 그리고, x축에 들어갈 값이 (날짜 등과같은) 존재할 경우 아래와 같은 .plot로 더 간단하게 그릴 수 있다. (Seaborn 보다 plt가 편하다.)

In [38]:

data.plot(figsize = (20,10))

Out[38]:

<AxesSubplot:xlabel='Order Date'>

추세를 예측하기 어려운 그림이다. 월 단위로 묶어서 다시 보자. .resample() 함수롤 사용한다.

In [40]:

data.resample('MS').mean()  # MS는 월별, 그리고 평균값 mean() 

Out[40]:

	Sales
Order Date
2014-01-01	480.194231
2014-02-01	367.931600
2014-03-01	857.291529
2014-04-01	567.488357
2014-05-01	432.049187
2014-06-01	695.059242
2014-07-01	601.169500
2014-08-01	457.521656
2014-09-01	992.353367
2014-10-01	769.015437
2014-11-01	980.221486
2014-12-01	1532.298325
2015-01-01	978.328467
2015-02-01	522.395667
2015-03-01	781.236437
2015-04-01	805.822962
2015-05-01	624.996700
2015-06-01	428.565500
2015-07-01	719.706316
2015-08-01	602.412012
2015-09-01	1382.790684
2015-10-01	632.980184
2015-11-01	1286.701354
2015-12-01	1049.355418
2016-01-01	508.182867
2016-02-01	356.868273
2016-03-01	609.575810
2016-04-01	695.373158
2016-05-01	687.265227
2016-06-01	816.910750
2016-07-01	768.736412
2016-08-01	734.307782
2016-09-01	1135.953371
2016-10-01	624.872474
2016-11-01	1271.345152
2016-12-01	1410.719808
2017-01-01	397.602133
2017-02-01	528.179800
2017-03-01	544.672240
2017-04-01	453.297905
2017-05-01	678.302328
2017-06-01	826.460291
2017-07-01	562.524857
2017-08-01	857.881889
2017-09-01	1209.508583
2017-10-01	875.362728
2017-11-01	1277.817759
2017-12-01	1256.298672

In [42]:

data = data.resample('MS').mean() # 최종적으로 사용할 data 완성 (할당해주자)

In [43]:

data.plot(figsize = (20,10))

Out[43]:

<AxesSubplot:xlabel='Order Date'>

In [44]:

from fbprophet import Prophet

In [45]:

data

Out[45]:

	Sales
Order Date
2014-01-01	480.194231
2014-02-01	367.931600
2014-03-01	857.291529
2014-04-01	567.488357
2014-05-01	432.049187
2014-06-01	695.059242
2014-07-01	601.169500
2014-08-01	457.521656
2014-09-01	992.353367
2014-10-01	769.015437
2014-11-01	980.221486
2014-12-01	1532.298325
2015-01-01	978.328467
2015-02-01	522.395667
2015-03-01	781.236437
2015-04-01	805.822962
2015-05-01	624.996700
2015-06-01	428.565500
2015-07-01	719.706316
2015-08-01	602.412012
2015-09-01	1382.790684
2015-10-01	632.980184
2015-11-01	1286.701354
2015-12-01	1049.355418
2016-01-01	508.182867
2016-02-01	356.868273
2016-03-01	609.575810
2016-04-01	695.373158
2016-05-01	687.265227
2016-06-01	816.910750
2016-07-01	768.736412
2016-08-01	734.307782
2016-09-01	1135.953371
2016-10-01	624.872474
2016-11-01	1271.345152
2016-12-01	1410.719808
2017-01-01	397.602133
2017-02-01	528.179800
2017-03-01	544.672240
2017-04-01	453.297905
2017-05-01	678.302328
2017-06-01	826.460291
2017-07-01	562.524857
2017-08-01	857.881889
2017-09-01	1209.508583
2017-10-01	875.362728
2017-11-01	1277.817759
2017-12-01	1256.298672

fbprophet을 사용하기 위해서는 몇몇 조건이 필요함으로 변경해야 한다. 이것을 알아보자. 컬럼이 ds 와 y 로 나타내야함.

In [51]:

df = data.reset_index()

In [52]:

df

Out[52]:

	Order Date	Sales
0	2014-01-01	480.194231
1	2014-02-01	367.931600
2	2014-03-01	857.291529
3	2014-04-01	567.488357
4	2014-05-01	432.049187
5	2014-06-01	695.059242
6	2014-07-01	601.169500
7	2014-08-01	457.521656
8	2014-09-01	992.353367
9	2014-10-01	769.015437
10	2014-11-01	980.221486
11	2014-12-01	1532.298325
12	2015-01-01	978.328467
13	2015-02-01	522.395667
14	2015-03-01	781.236437
15	2015-04-01	805.822962
16	2015-05-01	624.996700
17	2015-06-01	428.565500
18	2015-07-01	719.706316
19	2015-08-01	602.412012
20	2015-09-01	1382.790684
21	2015-10-01	632.980184
22	2015-11-01	1286.701354
23	2015-12-01	1049.355418
24	2016-01-01	508.182867
25	2016-02-01	356.868273
26	2016-03-01	609.575810
27	2016-04-01	695.373158
28	2016-05-01	687.265227
29	2016-06-01	816.910750
30	2016-07-01	768.736412
31	2016-08-01	734.307782
32	2016-09-01	1135.953371
33	2016-10-01	624.872474
34	2016-11-01	1271.345152
35	2016-12-01	1410.719808
36	2017-01-01	397.602133
37	2017-02-01	528.179800
38	2017-03-01	544.672240
39	2017-04-01	453.297905
40	2017-05-01	678.302328
41	2017-06-01	826.460291
42	2017-07-01	562.524857
43	2017-08-01	857.881889
44	2017-09-01	1209.508583
45	2017-10-01	875.362728
46	2017-11-01	1277.817759
47	2017-12-01	1256.298672

In [53]:

df.columns = ['ds','y']

In [54]:

 df

Out[54]:

	ds	y
0	2014-01-01	480.194231
1	2014-02-01	367.931600
2	2014-03-01	857.291529
3	2014-04-01	567.488357
4	2014-05-01	432.049187
5	2014-06-01	695.059242
6	2014-07-01	601.169500
7	2014-08-01	457.521656
8	2014-09-01	992.353367
9	2014-10-01	769.015437
10	2014-11-01	980.221486
11	2014-12-01	1532.298325
12	2015-01-01	978.328467
13	2015-02-01	522.395667
14	2015-03-01	781.236437
15	2015-04-01	805.822962
16	2015-05-01	624.996700
17	2015-06-01	428.565500
18	2015-07-01	719.706316
19	2015-08-01	602.412012
20	2015-09-01	1382.790684
21	2015-10-01	632.980184
22	2015-11-01	1286.701354
23	2015-12-01	1049.355418
24	2016-01-01	508.182867
25	2016-02-01	356.868273
26	2016-03-01	609.575810
27	2016-04-01	695.373158
28	2016-05-01	687.265227
29	2016-06-01	816.910750
30	2016-07-01	768.736412
31	2016-08-01	734.307782
32	2016-09-01	1135.953371
33	2016-10-01	624.872474
34	2016-11-01	1271.345152
35	2016-12-01	1410.719808
36	2017-01-01	397.602133
37	2017-02-01	528.179800
38	2017-03-01	544.672240
39	2017-04-01	453.297905
40	2017-05-01	678.302328
41	2017-06-01	826.460291
42	2017-07-01	562.524857
43	2017-08-01	857.881889
44	2017-09-01	1209.508583
45	2017-10-01	875.362728
46	2017-11-01	1277.817759
47	2017-12-01	1256.298672

In [55]:

model = Prophet()

In [57]:

model.fit(df)

INFO:fbprophet:Disabling weekly seasonality. Run prophet with weekly_seasonality=True to override this.
INFO:fbprophet:Disabling daily seasonality. Run prophet with daily_seasonality=True to override this.

Out[57]:

<fbprophet.forecaster.Prophet at 0x7f7983be3f98>

예측하고자하는 모델 범위 지정 .make_future_dataframe() 함수 사용

In [63]:

future = model.make_future_dataframe(periods = 12, freq = 'MS')   # 12개월 / 월 단위 리셋을 나타내는 파라미터 (periods = 12, freq = 'MS')

In [64]:

future

Out[64]:

	ds
0	2014-01-01
1	2014-02-01
2	2014-03-01
3	2014-04-01
4	2014-05-01
5	2014-06-01
6	2014-07-01
7	2014-08-01
8	2014-09-01
9	2014-10-01
10	2014-11-01
11	2014-12-01
12	2015-01-01
13	2015-02-01
14	2015-03-01
15	2015-04-01
16	2015-05-01
17	2015-06-01
18	2015-07-01
19	2015-08-01
20	2015-09-01
21	2015-10-01
22	2015-11-01
23	2015-12-01
24	2016-01-01
25	2016-02-01
26	2016-03-01
27	2016-04-01
28	2016-05-01
29	2016-06-01
30	2016-07-01
31	2016-08-01
32	2016-09-01
33	2016-10-01
34	2016-11-01
35	2016-12-01
36	2017-01-01
37	2017-02-01
38	2017-03-01
39	2017-04-01
40	2017-05-01
41	2017-06-01
42	2017-07-01
43	2017-08-01
44	2017-09-01
45	2017-10-01
46	2017-11-01
47	2017-12-01
48	2018-01-01
49	2018-02-01
50	2018-03-01
51	2018-04-01
52	2018-05-01
53	2018-06-01
54	2018-07-01
55	2018-08-01
56	2018-09-01
57	2018-10-01
58	2018-11-01
59	2018-12-01

In [65]:

model.predict(future)

Out[65]:

	ds	trend	yhat_lower	yhat_upper	trend_lower	trend_upper	additive_terms	additive_terms_lower	additive_terms_upper	yearly	yearly_lower	yearly_upper	multiplicative_terms	multiplicative_terms_lower	multiplicative_terms_upper	yhat
0	2014-01-01	726.057713	375.555739	695.295314	726.057713	726.057713	-190.685662	-190.685662	-190.685662	-190.685662	-190.685662	-190.685662	0.0	0.0	0.0	535.372051
1	2014-02-01	727.494023	293.399920	613.931087	727.494023	727.494023	-276.377703	-276.377703	-276.377703	-276.377703	-276.377703	-276.377703	0.0	0.0	0.0	451.116320
2	2014-03-01	728.791335	551.787401	875.526653	728.791335	728.791335	-22.389755	-22.389755	-22.389755	-22.389755	-22.389755	-22.389755	0.0	0.0	0.0	706.401580
3	2014-04-01	730.227645	479.070561	779.563943	730.227645	730.227645	-100.141158	-100.141158	-100.141158	-100.141158	-100.141158	-100.141158	0.0	0.0	0.0	630.086487
4	2014-05-01	731.617622	407.290288	727.947444	731.617622	731.617622	-160.815662	-160.815662	-160.815662	-160.815662	-160.815662	-160.815662	0.0	0.0	0.0	570.801960
5	2014-06-01	733.053932	464.110554	775.510263	733.053932	733.053932	-113.832366	-113.832366	-113.832366	-113.832366	-113.832366	-113.832366	0.0	0.0	0.0	619.221566
6	2014-07-01	734.443909	489.017230	815.461714	734.443909	734.443909	-81.335686	-81.335686	-81.335686	-81.335686	-81.335686	-81.335686	0.0	0.0	0.0	653.108224
7	2014-08-01	735.880219	451.040361	773.906672	735.880219	735.880219	-117.154950	-117.154950	-117.154950	-117.154950	-117.154950	-117.154950	0.0	0.0	0.0	618.725269
8	2014-09-01	737.316529	1041.500636	1351.531705	737.316529	737.316529	461.565905	461.565905	461.565905	461.565905	461.565905	461.565905	0.0	0.0	0.0	1198.882434
9	2014-10-01	738.706506	552.932184	858.826910	738.706506	738.706506	-28.820414	-28.820414	-28.820414	-28.820414	-28.820414	-28.820414	0.0	0.0	0.0	709.886092
10	2014-11-01	740.142816	1008.819647	1317.092508	740.142816	740.142816	431.055221	431.055221	431.055221	431.055221	431.055221	431.055221	0.0	0.0	0.0	1171.198037
11	2014-12-01	741.532794	1089.161899	1393.338664	741.532794	741.532794	501.882316	501.882316	501.882316	501.882316	501.882316	501.882316	0.0	0.0	0.0	1243.415110
12	2015-01-01	742.969103	454.173451	767.734489	742.969103	742.969103	-130.055125	-130.055125	-130.055125	-130.055125	-130.055125	-130.055125	0.0	0.0	0.0	612.913979
13	2015-02-01	744.405413	254.091386	577.236965	744.405413	744.405413	-341.079140	-341.079140	-341.079140	-341.079140	-341.079140	-341.079140	0.0	0.0	0.0	403.326273
14	2015-03-01	745.702725	642.430576	951.623020	745.702725	745.702725	46.620746	46.620746	46.620746	46.620746	46.620746	46.620746	0.0	0.0	0.0	792.323472
15	2015-04-01	747.139035	540.239706	845.960559	747.139035	747.139035	-53.528569	-53.528569	-53.528569	-53.528569	-53.528569	-53.528569	0.0	0.0	0.0	693.610467
16	2015-05-01	748.529013	417.522498	714.561871	748.529013	748.529013	-184.631339	-184.631339	-184.631339	-184.631339	-184.631339	-184.631339	0.0	0.0	0.0	563.897674
17	2015-06-01	749.965323	395.877547	704.827872	749.965323	749.965323	-207.846148	-207.846148	-207.846148	-207.846148	-207.846148	-207.846148	0.0	0.0	0.0	542.119174
18	2015-07-01	751.355300	546.422761	852.998396	751.355300	751.355300	-48.294657	-48.294657	-48.294657	-48.294657	-48.294657	-48.294657	0.0	0.0	0.0	703.060643
19	2015-08-01	752.791609	448.121103	756.607455	752.791609	752.791609	-152.742582	-152.742582	-152.742582	-152.742582	-152.742582	-152.742582	0.0	0.0	0.0	600.049027
20	2015-09-01	754.227919	1149.095495	1461.573191	754.227919	754.227919	551.713042	551.713042	551.713042	551.713042	551.713042	551.713042	0.0	0.0	0.0	1305.940961
21	2015-10-01	755.617896	588.050589	893.732664	755.617896	755.617896	-13.792198	-13.792198	-13.792198	-13.792198	-13.792198	-13.792198	0.0	0.0	0.0	741.825698
22	2015-11-01	757.054205	1006.264367	1330.522230	757.054205	757.054205	413.144741	413.144741	413.144741	413.144741	413.144741	413.144741	0.0	0.0	0.0	1170.198947
23	2015-12-01	758.444182	1024.755884	1326.386146	758.444182	758.444182	417.365653	417.365653	417.365653	417.365653	417.365653	417.365653	0.0	0.0	0.0	1175.809835
24	2016-01-01	759.880491	533.801284	845.522689	759.880491	759.880491	-68.554067	-68.554067	-68.554067	-68.554067	-68.554067	-68.554067	0.0	0.0	0.0	691.326424
25	2016-02-01	761.316801	191.493274	514.488624	761.316801	761.316801	-405.014691	-405.014691	-405.014691	-405.014691	-405.014691	-405.014691	0.0	0.0	0.0	356.302110
26	2016-03-01	762.660445	465.376362	752.621365	762.660445	762.660445	-157.661809	-157.661809	-157.661809	-157.661809	-157.661809	-157.661809	0.0	0.0	0.0	604.998637
27	2016-04-01	764.096755	397.607966	720.010866	764.096755	764.096755	-197.922888	-197.922888	-197.922888	-197.922888	-197.922888	-197.922888	0.0	0.0	0.0	566.173867
28	2016-05-01	765.486732	493.020424	795.877936	765.486732	765.486732	-119.761707	-119.761707	-119.761707	-119.761707	-119.761707	-119.761707	0.0	0.0	0.0	645.725025
29	2016-06-01	766.923041	676.173435	994.722300	766.923041	766.923041	75.113515	75.113515	75.113515	75.113515	75.113515	75.113515	0.0	0.0	0.0	842.036556
30	2016-07-01	768.313018	472.799304	790.070746	768.313018	768.313018	-143.219273	-143.219273	-143.219273	-143.219273	-143.219273	-143.219273	0.0	0.0	0.0	625.093745
31	2016-08-01	769.749327	584.580146	889.021341	769.749327	769.749327	-41.282256	-41.282256	-41.282256	-41.282256	-41.282256	-41.282256	0.0	0.0	0.0	728.467072
32	2016-09-01	771.185637	887.179005	1204.828346	771.185637	771.185637	282.865686	282.865686	282.865686	282.865686	282.865686	282.865686	0.0	0.0	0.0	1054.051323
33	2016-10-01	772.575614	545.914732	860.391636	772.575614	772.575614	-63.840536	-63.840536	-63.840536	-63.840536	-63.840536	-63.840536	0.0	0.0	0.0	708.735078
34	2016-11-01	774.011923	1085.839458	1392.254970	774.011923	774.011923	461.298523	461.298523	461.298523	461.298523	461.298523	461.298523	0.0	0.0	0.0	1235.310446
35	2016-12-01	775.401900	1287.888860	1602.762277	775.401900	775.401900	672.698002	672.698002	672.698002	672.698002	672.698002	672.698002	0.0	0.0	0.0	1448.099902
36	2017-01-01	776.838209	369.980424	683.107377	776.838209	776.838209	-250.426116	-250.426116	-250.426116	-250.426116	-250.426116	-250.426116	0.0	0.0	0.0	526.412094
37	2017-02-01	778.274519	408.308622	717.390143	778.274519	778.274519	-210.957398	-210.957398	-210.957398	-210.957398	-210.957398	-210.957398	0.0	0.0	0.0	567.317120
38	2017-03-01	779.571831	535.357796	850.592519	779.571831	779.571831	-90.513443	-90.513443	-90.513443	-90.513443	-90.513443	-90.513443	0.0	0.0	0.0	689.058388
39	2017-04-01	781.008140	477.439584	785.557238	781.008140	781.008140	-148.296501	-148.296501	-148.296501	-148.296501	-148.296501	-148.296501	0.0	0.0	0.0	632.711639
40	2017-05-01	782.398117	486.007271	810.160435	782.398117	782.398117	-139.178722	-139.178722	-139.178722	-139.178722	-139.178722	-139.178722	0.0	0.0	0.0	643.219395
41	2017-06-01	783.834426	603.733179	922.225522	783.834426	783.834426	-19.474492	-19.474492	-19.474492	-19.474492	-19.474492	-19.474492	0.0	0.0	0.0	764.359934
42	2017-07-01	785.224403	517.022563	829.998891	785.224403	785.224403	-112.977624	-112.977624	-112.977624	-112.977624	-112.977624	-112.977624	0.0	0.0	0.0	672.246779
43	2017-08-01	786.660713	553.528488	866.557563	786.660713	786.660713	-79.999841	-79.999841	-79.999841	-79.999841	-79.999841	-79.999841	0.0	0.0	0.0	706.660871
44	2017-09-01	788.097022	996.618534	1316.692991	788.097022	788.097022	371.914804	371.914804	371.914804	371.914804	371.914804	371.914804	0.0	0.0	0.0	1160.011826
45	2017-10-01	789.486999	585.531905	888.099082	789.486999	789.486999	-45.516303	-45.516303	-45.516303	-45.516303	-45.516303	-45.516303	0.0	0.0	0.0	743.970696
46	2017-11-01	790.923308	1080.626366	1400.663856	790.923308	790.923308	447.117002	447.117002	447.117002	447.117002	447.117002	447.117002	0.0	0.0	0.0	1238.040310
47	2017-12-01	792.313285	1221.074372	1538.605895	792.313285	792.313285	587.025701	587.025701	587.025701	587.025701	587.025701	587.025701	0.0	0.0	0.0	1379.338986
48	2018-01-01	793.749595	448.163047	756.313780	793.749595	793.749595	-190.685662	-190.685662	-190.685662	-190.685662	-190.685662	-190.685662	0.0	0.0	0.0	603.063933
49	2018-02-01	795.185904	360.828186	683.291792	795.185904	795.185905	-276.377703	-276.377703	-276.377703	-276.377703	-276.377703	-276.377703	0.0	0.0	0.0	518.808201
50	2018-03-01	796.483216	614.640162	925.996570	796.483215	796.483217	-22.389755	-22.389755	-22.389755	-22.389755	-22.389755	-22.389755	0.0	0.0	0.0	774.093461
51	2018-04-01	797.919525	534.060372	852.194196	797.919524	797.919527	-100.141158	-100.141158	-100.141158	-100.141158	-100.141158	-100.141158	0.0	0.0	0.0	697.778367
52	2018-05-01	799.309502	474.410237	781.590787	799.309500	799.309505	-160.815662	-160.815662	-160.815662	-160.815662	-160.815662	-160.815662	0.0	0.0	0.0	638.493840
53	2018-06-01	800.745812	537.799138	834.481771	800.745808	800.745815	-113.832366	-113.832366	-113.832366	-113.832366	-113.832366	-113.832366	0.0	0.0	0.0	686.913446
54	2018-07-01	802.135789	561.579231	875.890879	802.135784	802.135793	-81.335686	-81.335686	-81.335686	-81.335686	-81.335686	-81.335686	0.0	0.0	0.0	720.800103
55	2018-08-01	803.572098	532.489762	837.208640	803.572093	803.572104	-117.154950	-117.154950	-117.154950	-117.154950	-117.154950	-117.154950	0.0	0.0	0.0	686.417148
56	2018-09-01	805.008408	1123.623993	1429.923019	805.008401	805.008415	461.565905	461.565905	461.565905	461.565905	461.565905	461.565905	0.0	0.0	0.0	1266.574313
57	2018-10-01	806.398384	623.089438	930.323047	806.398376	806.398393	-28.820414	-28.820414	-28.820414	-28.820414	-28.820414	-28.820414	0.0	0.0	0.0	777.577970
58	2018-11-01	807.834694	1083.933747	1395.291076	807.834685	807.834703	431.055221	431.055221	431.055221	431.055221	431.055221	431.055221	0.0	0.0	0.0	1238.889915
59	2018-12-01	809.224671	1150.847123	1475.396456	809.224661	809.224681	501.882316	501.882316	501.882316	501.882316	501.882316	501.882316	0.0	0.0	0.0	1311.106987

In [69]:

pred = model.predict(future)

In [70]:

pred = pred[['ds', 'yhat', 'yhat_lower', 'yhat_upper']]

In [71]:

pred

Out[71]:

	ds	yhat	yhat_lower	yhat_upper
0	2014-01-01	535.372051	378.524766	688.636959
1	2014-02-01	451.116320	293.796833	610.251748
2	2014-03-01	706.401580	543.923578	859.173669
3	2014-04-01	630.086487	466.414631	788.518361
4	2014-05-01	570.801960	418.062888	723.688846
5	2014-06-01	619.221566	473.588452	782.257305
6	2014-07-01	653.108224	489.390112	817.836053
7	2014-08-01	618.725269	465.640830	777.228368
8	2014-09-01	1198.882434	1032.947053	1357.184833
9	2014-10-01	709.886092	557.926691	863.095101
10	2014-11-01	1171.198037	1000.900573	1326.257329
11	2014-12-01	1243.415110	1078.151134	1399.100457
12	2015-01-01	612.913979	459.295261	774.310468
13	2015-02-01	403.326273	244.878685	557.974096
14	2015-03-01	792.323472	629.197645	949.559294
15	2015-04-01	693.610467	552.540290	849.330312
16	2015-05-01	563.897674	407.523113	717.615703
17	2015-06-01	542.119174	377.671689	698.468811
18	2015-07-01	703.060643	554.042171	863.980917
19	2015-08-01	600.049027	441.376748	743.052218
20	2015-09-01	1305.940961	1152.713902	1469.983934
21	2015-10-01	741.825698	587.322735	905.992719
22	2015-11-01	1170.198947	1016.291767	1332.152695
23	2015-12-01	1175.809835	1019.107049	1332.031156
24	2016-01-01	691.326424	536.685486	841.037102
25	2016-02-01	356.302110	192.462275	509.339288
26	2016-03-01	604.998637	451.336717	751.128700
27	2016-04-01	566.173867	415.777235	716.142581
28	2016-05-01	645.725025	495.090228	807.995724
29	2016-06-01	842.036556	684.175196	1001.981301
30	2016-07-01	625.093745	460.043929	785.621332
31	2016-08-01	728.467072	575.005988	891.436275
32	2016-09-01	1054.051323	900.880527	1211.208162
33	2016-10-01	708.735078	555.187877	873.956714
34	2016-11-01	1235.310446	1090.615284	1390.800265
35	2016-12-01	1448.099902	1295.800649	1607.336201
36	2017-01-01	526.412094	376.730524	684.014411
37	2017-02-01	567.317120	409.149813	730.286661
38	2017-03-01	689.058388	531.694572	842.986690
39	2017-04-01	632.711639	476.111002	792.837900
40	2017-05-01	643.219395	490.185802	795.416521
41	2017-06-01	764.359934	611.237269	919.585267
42	2017-07-01	672.246779	513.125091	828.246671
43	2017-08-01	706.660871	546.766929	864.654936
44	2017-09-01	1160.011826	1005.427189	1325.637425
45	2017-10-01	743.970696	578.941466	907.667671
46	2017-11-01	1238.040310	1082.147207	1392.479155
47	2017-12-01	1379.338986	1216.319883	1536.014510
48	2018-01-01	603.063933	447.464202	765.737086
49	2018-02-01	518.808201	353.375332	677.391527
50	2018-03-01	774.093461	615.274334	930.740848
51	2018-04-01	697.778367	537.821410	861.134434
52	2018-05-01	638.493840	477.662807	792.591736
53	2018-06-01	686.913446	526.867767	830.127164
54	2018-07-01	720.800103	555.966558	873.713180
55	2018-08-01	686.417148	531.007182	852.261733
56	2018-09-01	1266.574313	1098.857803	1423.457812
57	2018-10-01	777.577970	620.248696	936.594107
58	2018-11-01	1238.889915	1083.628999	1385.253346
59	2018-12-01	1311.106987	1161.211924	1467.705491

In [73]:

model.plot(pred)
plt.show()

Train ,Test data 셋을 나누어보자.

In [74]:

data.head()

Out[74]:

	Sales
Order Date
2014-01-01	480.194231
2014-02-01	367.931600
2014-03-01	857.291529
2014-04-01	567.488357
2014-05-01	432.049187

In [77]:

data[:'2016']   # 2016년 까지의 data만 불러오기

Out[77]:

	Sales
Order Date
2014-01-01	480.194231
2014-02-01	367.931600
2014-03-01	857.291529
2014-04-01	567.488357
2014-05-01	432.049187
2014-06-01	695.059242
2014-07-01	601.169500
2014-08-01	457.521656
2014-09-01	992.353367
2014-10-01	769.015437
2014-11-01	980.221486
2014-12-01	1532.298325
2015-01-01	978.328467
2015-02-01	522.395667
2015-03-01	781.236437
2015-04-01	805.822962
2015-05-01	624.996700
2015-06-01	428.565500
2015-07-01	719.706316
2015-08-01	602.412012
2015-09-01	1382.790684
2015-10-01	632.980184
2015-11-01	1286.701354
2015-12-01	1049.355418
2016-01-01	508.182867
2016-02-01	356.868273
2016-03-01	609.575810
2016-04-01	695.373158
2016-05-01	687.265227
2016-06-01	816.910750
2016-07-01	768.736412
2016-08-01	734.307782
2016-09-01	1135.953371
2016-10-01	624.872474
2016-11-01	1271.345152
2016-12-01	1410.719808

2016까지의 데이터를 훈련데이터로 쓰고, 2017년 데이터는 쌩으로 데이터로 사용하고자 한다.

In [78]:

new_df = data[:'2016']

In [79]:

new_df = new_df.reset_index()

In [80]:

new_df.columns = ['ds', 'y']

In [81]:

new_df

Out[81]:

	ds	y
0	2014-01-01	480.194231
1	2014-02-01	367.931600
2	2014-03-01	857.291529
3	2014-04-01	567.488357
4	2014-05-01	432.049187
5	2014-06-01	695.059242
6	2014-07-01	601.169500
7	2014-08-01	457.521656
8	2014-09-01	992.353367
9	2014-10-01	769.015437
10	2014-11-01	980.221486
11	2014-12-01	1532.298325
12	2015-01-01	978.328467
13	2015-02-01	522.395667
14	2015-03-01	781.236437
15	2015-04-01	805.822962
16	2015-05-01	624.996700
17	2015-06-01	428.565500
18	2015-07-01	719.706316
19	2015-08-01	602.412012
20	2015-09-01	1382.790684
21	2015-10-01	632.980184
22	2015-11-01	1286.701354
23	2015-12-01	1049.355418
24	2016-01-01	508.182867
25	2016-02-01	356.868273
26	2016-03-01	609.575810
27	2016-04-01	695.373158
28	2016-05-01	687.265227
29	2016-06-01	816.910750
30	2016-07-01	768.736412
31	2016-08-01	734.307782
32	2016-09-01	1135.953371
33	2016-10-01	624.872474
34	2016-11-01	1271.345152
35	2016-12-01	1410.719808

In [122]:

model = Prophet()

In [123]:

model.fit(new_df)

INFO:fbprophet:Disabling weekly seasonality. Run prophet with weekly_seasonality=True to override this.
INFO:fbprophet:Disabling daily seasonality. Run prophet with daily_seasonality=True to override this.

Out[123]:

<fbprophet.forecaster.Prophet at 0x7f7983014a20>

In [124]:

future = model.make_future_dataframe(periods = 12, freq = 'MS') # 예상할 범위 지정

In [125]:

pred = model.predict(future)

In [126]:

pred

Out[126]:

	ds	trend	yhat_lower	yhat_upper	trend_lower	trend_upper	additive_terms	additive_terms_lower	additive_terms_upper	yearly	yearly_lower	yearly_upper	multiplicative_terms	multiplicative_terms_lower	multiplicative_terms_upper	yhat
0	2014-01-01	681.588219	489.820917	780.550959	681.588219	681.588219	-47.448838	-47.448838	-47.448838	-47.448838	-47.448838	-47.448838	0.0	0.0	0.0	634.139381
1	2014-02-01	685.636335	263.817165	551.384047	685.636335	685.636335	-274.076247	-274.076247	-274.076247	-274.076247	-274.076247	-274.076247	0.0	0.0	0.0	411.560088
2	2014-03-01	689.292697	590.943677	877.179408	689.292697	689.292697	42.432009	42.432009	42.432009	42.432009	42.432009	42.432009	0.0	0.0	0.0	731.724706
3	2014-04-01	693.340812	506.537667	808.965787	693.340812	693.340812	-34.565068	-34.565068	-34.565068	-34.565068	-34.565068	-34.565068	0.0	0.0	0.0	658.775745
4	2014-05-01	697.258343	380.459041	691.741707	697.258343	697.258343	-165.311506	-165.311506	-165.311506	-165.311506	-165.311506	-165.311506	0.0	0.0	0.0	531.946837
5	2014-06-01	701.306459	420.211069	723.366012	701.306459	701.306459	-133.386474	-133.386474	-133.386474	-133.386474	-133.386474	-133.386474	0.0	0.0	0.0	567.919984
6	2014-07-01	705.223990	497.611700	804.471583	705.223990	705.223990	-58.042923	-58.042923	-58.042923	-58.042923	-58.042923	-58.042923	0.0	0.0	0.0	647.181067
7	2014-08-01	709.272106	383.683725	693.950098	709.272106	709.272106	-171.626465	-171.626465	-171.626465	-171.626465	-171.626465	-171.626465	0.0	0.0	0.0	537.645640
8	2014-09-01	713.320221	1001.869125	1292.353941	713.320221	713.320221	436.759246	436.759246	436.759246	436.759246	436.759246	436.759246	0.0	0.0	0.0	1150.079467
9	2014-10-01	717.237752	500.833488	795.834748	717.237752	717.237752	-76.340818	-76.340818	-76.340818	-76.340818	-76.340818	-76.340818	0.0	0.0	0.0	640.896934
10	2014-11-01	721.285868	982.698246	1280.944120	721.285868	721.285868	414.478990	414.478990	414.478990	414.478990	414.478990	414.478990	0.0	0.0	0.0	1135.764858
11	2014-12-01	725.203399	1114.032354	1407.626220	725.203399	725.203399	535.324793	535.324793	535.324793	535.324793	535.324793	535.324793	0.0	0.0	0.0	1260.528192
12	2015-01-01	729.251514	507.046250	802.452572	729.251514	729.251514	-74.004639	-74.004639	-74.004639	-74.004639	-74.004639	-74.004639	0.0	0.0	0.0	655.246875
13	2015-02-01	733.299630	270.205334	551.769970	733.299630	733.299630	-318.385936	-318.385936	-318.385936	-318.385936	-318.385936	-318.385936	0.0	0.0	0.0	414.913693
14	2015-03-01	736.955992	717.461038	1024.298701	736.955992	736.955992	135.207544	135.207544	135.207544	135.207544	135.207544	135.207544	0.0	0.0	0.0	872.163536
15	2015-04-01	741.004093	599.434707	893.316256	741.004093	741.004093	10.925079	10.925079	10.925079	10.925079	10.925079	10.925079	0.0	0.0	0.0	751.929172
16	2015-05-01	744.921610	423.294725	718.730926	744.921610	744.921610	-178.100732	-178.100732	-178.100732	-178.100732	-178.100732	-178.100732	0.0	0.0	0.0	566.820879
17	2015-06-01	748.969712	376.036629	685.465799	748.969712	748.969712	-228.203393	-228.203393	-228.203393	-228.203393	-228.203393	-228.203393	0.0	0.0	0.0	520.766319
18	2015-07-01	752.887229	552.088154	840.950943	752.887229	752.887229	-58.585019	-58.585019	-58.585019	-58.585019	-58.585019	-58.585019	0.0	0.0	0.0	694.302210
19	2015-08-01	756.935331	408.764592	705.574247	756.935331	756.935331	-203.124853	-203.124853	-203.124853	-203.124853	-203.124853	-203.124853	0.0	0.0	0.0	553.810477
20	2015-09-01	760.983432	1138.933806	1429.417552	760.983432	760.983432	522.378502	522.378502	522.378502	522.378502	522.378502	522.378502	0.0	0.0	0.0	1283.361934
21	2015-10-01	764.900950	576.829257	872.963723	764.900950	764.900950	-41.132905	-41.132905	-41.132905	-41.132905	-41.132905	-41.132905	0.0	0.0	0.0	723.768045
22	2015-11-01	768.949051	1034.928202	1353.467806	768.949051	768.949051	420.101231	420.101231	420.101231	420.101231	420.101231	420.101231	0.0	0.0	0.0	1189.050282
23	2015-12-01	772.866568	1085.980872	1387.241224	772.866568	772.866568	464.683096	464.683096	464.683096	464.683096	464.683096	464.683096	0.0	0.0	0.0	1237.549665
24	2016-01-01	776.914670	533.434123	824.662898	776.914670	776.914670	-99.647965	-99.647965	-99.647965	-99.647965	-99.647965	-99.647965	0.0	0.0	0.0	677.266704
25	2016-02-01	780.962771	269.521829	570.783075	780.962771	780.962771	-360.434238	-360.434238	-360.434238	-360.434238	-360.434238	-360.434238	0.0	0.0	0.0	420.528534
26	2016-03-01	784.749705	496.867187	793.121656	784.749705	784.749705	-140.486697	-140.486697	-140.486697	-140.486697	-140.486697	-140.486697	0.0	0.0	0.0	644.263007
27	2016-04-01	788.797806	529.030838	811.756415	788.797806	788.797806	-130.923223	-130.923223	-130.923223	-130.923223	-130.923223	-130.923223	0.0	0.0	0.0	657.874584
28	2016-05-01	792.715324	500.902636	785.870123	792.715324	792.715324	-147.271749	-147.271749	-147.271749	-147.271749	-147.271749	-147.271749	0.0	0.0	0.0	645.443575
29	2016-06-01	796.763425	698.191855	986.436287	796.763425	796.763425	55.078814	55.078814	55.078814	55.078814	55.078814	55.078814	0.0	0.0	0.0	851.842239
30	2016-07-01	800.680943	599.818969	900.783295	800.680943	800.680943	-52.647345	-52.647345	-52.647345	-52.647345	-52.647345	-52.647345	0.0	0.0	0.0	748.033598
31	2016-08-01	804.729044	564.562668	854.640135	804.729044	804.729044	-102.089978	-102.089978	-102.089978	-102.089978	-102.089978	-102.089978	0.0	0.0	0.0	702.639066
32	2016-09-01	808.777146	935.301896	1230.359833	808.777146	808.777146	268.734098	268.734098	268.734098	268.734098	268.734098	268.734098	0.0	0.0	0.0	1077.511244
33	2016-10-01	812.694664	501.744256	798.487191	812.694664	812.694664	-150.546754	-150.546754	-150.546754	-150.546754	-150.546754	-150.546754	0.0	0.0	0.0	662.147909
34	2016-11-01	816.742765	1068.965055	1360.526758	816.742765	816.742765	396.487662	396.487662	396.487662	396.487662	396.487662	396.487662	0.0	0.0	0.0	1213.230427
35	2016-12-01	820.660283	1333.386531	1652.766966	820.660283	820.660283	673.686476	673.686476	673.686476	673.686476	673.686476	673.686476	0.0	0.0	0.0	1494.346758
36	2017-01-01	824.708384	654.628881	948.501143	824.708384	824.708385	-20.053946	-20.053946	-20.053946	-20.053946	-20.053946	-20.053946	0.0	0.0	0.0	804.654438
37	2017-02-01	828.756486	451.673740	739.229452	828.756484	828.756487	-227.549968	-227.549968	-227.549968	-227.549968	-227.549968	-227.549968	0.0	0.0	0.0	601.206518
38	2017-03-01	832.412835	647.384793	934.031460	832.412832	832.412839	-49.507766	-49.507766	-49.507766	-49.507766	-49.507766	-49.507766	0.0	0.0	0.0	782.905069
39	2017-04-01	836.460937	615.754206	904.336383	836.460931	836.460943	-81.875639	-81.875639	-81.875639	-81.875639	-81.875639	-81.875639	0.0	0.0	0.0	754.585298
40	2017-05-01	840.378455	536.074714	837.877014	840.378446	840.378463	-155.025651	-155.025651	-155.025651	-155.025651	-155.025651	-155.025651	0.0	0.0	0.0	685.352803
41	2017-06-01	844.426556	649.635087	939.254091	844.426544	844.426568	-38.918902	-38.918902	-38.918902	-38.918902	-38.918902	-38.918902	0.0	0.0	0.0	805.507655
42	2017-07-01	848.344074	650.086358	939.639765	848.344058	848.344089	-56.050359	-56.050359	-56.050359	-56.050359	-56.050359	-56.050359	0.0	0.0	0.0	792.293715
43	2017-08-01	852.392175	565.438941	866.059661	852.392157	852.392194	-137.943755	-137.943755	-137.943755	-137.943755	-137.943755	-137.943755	0.0	0.0	0.0	714.448421
44	2017-09-01	856.440277	1061.035048	1368.132822	856.440253	856.440299	352.175705	352.175705	352.175705	352.175705	352.175705	352.175705	0.0	0.0	0.0	1208.615981
45	2017-10-01	860.357794	599.915070	890.540836	860.357767	860.357820	-112.835925	-112.835925	-112.835925	-112.835925	-112.835925	-112.835925	0.0	0.0	0.0	747.521869
46	2017-11-01	864.405896	1131.166407	1424.700197	864.405865	864.405927	406.605191	406.605191	406.605191	406.605191	406.605191	406.605191	0.0	0.0	0.0	1271.011087
47	2017-12-01	868.323413	1333.490239	1617.829474	868.323378	868.323449	605.025364	605.025364	605.025364	605.025364	605.025364	605.025364	0.0	0.0	0.0	1473.348777

In [130]:

pred = pred[['ds', 'yhat']]

In [131]:

pred.set_index('ds', inplace=True)

In [132]:

pred = pred['2017':]

In [121]:

pred

Out[121]:

	yhat
ds
2017-01-01	804.654438
2017-02-01	601.206518
2017-03-01	782.905069
2017-04-01	754.585298
2017-05-01	685.352803
2017-06-01	805.507655
2017-07-01	792.293715
2017-08-01	714.448421
2017-09-01	1208.615981
2017-10-01	747.521869
2017-11-01	1271.011087
2017-12-01	1473.348777

In [99]:

data

Out[99]:

	Sales
Order Date
2014-01-01	480.194231
2014-02-01	367.931600
2014-03-01	857.291529
2014-04-01	567.488357
2014-05-01	432.049187
2014-06-01	695.059242
2014-07-01	601.169500
2014-08-01	457.521656
2014-09-01	992.353367
2014-10-01	769.015437
2014-11-01	980.221486
2014-12-01	1532.298325
2015-01-01	978.328467
2015-02-01	522.395667
2015-03-01	781.236437
2015-04-01	805.822962
2015-05-01	624.996700
2015-06-01	428.565500
2015-07-01	719.706316
2015-08-01	602.412012
2015-09-01	1382.790684
2015-10-01	632.980184
2015-11-01	1286.701354
2015-12-01	1049.355418
2016-01-01	508.182867
2016-02-01	356.868273
2016-03-01	609.575810
2016-04-01	695.373158
2016-05-01	687.265227
2016-06-01	816.910750
2016-07-01	768.736412
2016-08-01	734.307782
2016-09-01	1135.953371
2016-10-01	624.872474
2016-11-01	1271.345152
2016-12-01	1410.719808
2017-01-01	397.602133
2017-02-01	528.179800
2017-03-01	544.672240
2017-04-01	453.297905
2017-05-01	678.302328
2017-06-01	826.460291
2017-07-01	562.524857
2017-08-01	857.881889
2017-09-01	1209.508583
2017-10-01	875.362728
2017-11-01	1277.817759
2017-12-01	1256.298672

In [100]:

ax = data.plot()

In [102]:

ax = data.plot()
pred.plot(ax=ax, figsize=(20,10))    # ax plot에 pred plot를 겹처서 그리게 하는 파라미터를 추가한다. 

Out[102]:

<AxesSubplot:xlabel='ds'>

파란색은 실제 Sales의 값이며, 주황색은 우리가 예측한 값이다. 2017년 초반에는 차이가 좀 있지만, 전체 적인 트렌드는 비슷한것을 확인 할 수 있다.

3. AutoRegressive 알고리즘을 통한 예측 (성능은 prphet보다 떨어진다.)¶

In [159]:

data = pd.read_excel('./data/Superstore.xls', index_col=0)

In [160]:

data['Category'].unique()

Out[160]:

array(['Furniture', 'Office Supplies', 'Technology'], dtype=object)

.unique() 함수를 사용하여 Category의 종류를 확인해 볼 수 있다.

In [161]:

data['Category'].value_counts()

Out[161]:

Office Supplies    6026
Furniture          2121
Technology         1847
Name: Category, dtype: int64

.value_counts() 함수를 사용하여 Category의 종류와 수량을 확인 할 수 있다.

In [162]:

data['Category'] == 'Furniture'  ## 불리안 True or False 확인

Out[162]:

Row ID
1        True
2        True
3       False
4        True
5       False
        ...  
9990     True
9991     True
9992    False
9993    False
9994    False
Name: Category, Length: 9994, dtype: bool

In [163]:

data.head() # 위 코드와 대조해서 확인 해본다. 

Out[163]:

	Order ID	Order Date	Ship Date	Ship Mode	Customer ID	Customer Name	Segment	Country	City	State	Postal Code	Region	Product ID	Category	Sub-Category	Product Name	Sales	Quantity	Discount	Profit
Row ID
1	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-BO-10001798	Furniture	Bookcases	Bush Somerset Collection Bookcase	261.9600	2	0.00	41.9136
2	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-CH-10000454	Furniture	Chairs	Hon Deluxe Fabric Upholstered Stacking Chairs,...	731.9400	3	0.00	219.5820
3	CA-2016-138688	2016-06-12	2016-06-16	Second Class	DV-13045	Darrin Van Huff	Corporate	United States	Los Angeles	California	90036	West	OFF-LA-10000240	Office Supplies	Labels	Self-Adhesive Address Labels for Typewriters b...	14.6200	2	0.00	6.8714
4	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	FUR-TA-10000577	Furniture	Tables	Bretford CR4500 Series Slim Rectangular Table	957.5775	5	0.45	-383.0310
5	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	OFF-ST-10000760	Office Supplies	Storage	Eldon Fold 'N Roll Cart System	22.3680	2	0.20	2.5164

In [164]:

data[data['Category'] == 'Furniture' ] # data['Category'] == 'Furniture' 을 data[]에 넣어 주면 True만 불러와주는 기능을 한다.

Out[164]:

	Order ID	Order Date	Ship Date	Ship Mode	Customer ID	Customer Name	Segment	Country	City	State	Postal Code	Region	Product ID	Category	Sub-Category	Product Name	Sales	Quantity	Discount	Profit
Row ID
1	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-BO-10001798	Furniture	Bookcases	Bush Somerset Collection Bookcase	261.9600	2	0.00	41.9136
2	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-CH-10000454	Furniture	Chairs	Hon Deluxe Fabric Upholstered Stacking Chairs,...	731.9400	3	0.00	219.5820
4	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	FUR-TA-10000577	Furniture	Tables	Bretford CR4500 Series Slim Rectangular Table	957.5775	5	0.45	-383.0310
6	CA-2014-115812	2014-06-09	2014-06-14	Standard Class	BH-11710	Brosina Hoffman	Consumer	United States	Los Angeles	California	90032	West	FUR-FU-10001487	Furniture	Furnishings	Eldon Expressions Wood and Plastic Desk Access...	48.8600	7	0.00	14.1694
11	CA-2014-115812	2014-06-09	2014-06-14	Standard Class	BH-11710	Brosina Hoffman	Consumer	United States	Los Angeles	California	90032	West	FUR-TA-10001539	Furniture	Tables	Chromcraft Rectangular Conference Tables	1706.1840	9	0.20	85.3092
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
9963	CA-2015-168088	2015-03-19	2015-03-22	First Class	CM-12655	Corinna Mitchell	Home Office	United States	Houston	Texas	77041	Central	FUR-BO-10004218	Furniture	Bookcases	Bush Heritage Pine Collection 5-Shelf Bookcase...	383.4656	4	0.32	-67.6704
9965	CA-2016-146374	2016-12-05	2016-12-10	Second Class	HE-14800	Harold Engle	Corporate	United States	Newark	Delaware	19711	East	FUR-FU-10002671	Furniture	Furnishings	Electrix 20W Halogen Replacement Bulb for Zoom...	13.4000	1	0.00	6.4320
9981	US-2015-151435	2015-09-06	2015-09-09	Second Class	SW-20455	Shaun Weien	Consumer	United States	Lafayette	Louisiana	70506	South	FUR-TA-10001039	Furniture	Tables	KI Adjustable-Height Table	85.9800	1	0.00	22.3548
9990	CA-2014-110422	2014-01-21	2014-01-23	Second Class	TB-21400	Tom Boeckenhauer	Consumer	United States	Miami	Florida	33180	South	FUR-FU-10001889	Furniture	Furnishings	Ultra Door Pull Handle	25.2480	3	0.20	4.1028
9991	CA-2017-121258	2017-02-26	2017-03-03	Standard Class	DB-13060	Dave Brooks	Consumer	United States	Costa Mesa	California	92627	West	FUR-FU-10000747	Furniture	Furnishings	Tenex B1-RE Series Chair Mats for Low Pile Car...	91.9600	2	0.00	15.6332

2121 rows × 20 columns

In [165]:

data = data[data['Category'] == 'Furniture']  # data에 할당

In [166]:

data.head()

Out[166]:

	Order ID	Order Date	Ship Date	Ship Mode	Customer ID	Customer Name	Segment	Country	City	State	Postal Code	Region	Product ID	Category	Sub-Category	Product Name	Sales	Quantity	Discount	Profit
Row ID
1	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-BO-10001798	Furniture	Bookcases	Bush Somerset Collection Bookcase	261.9600	2	0.00	41.9136
2	CA-2016-152156	2016-11-08	2016-11-11	Second Class	CG-12520	Claire Gute	Consumer	United States	Henderson	Kentucky	42420	South	FUR-CH-10000454	Furniture	Chairs	Hon Deluxe Fabric Upholstered Stacking Chairs,...	731.9400	3	0.00	219.5820
4	US-2015-108966	2015-10-11	2015-10-18	Standard Class	SO-20335	Sean O'Donnell	Consumer	United States	Fort Lauderdale	Florida	33311	South	FUR-TA-10000577	Furniture	Tables	Bretford CR4500 Series Slim Rectangular Table	957.5775	5	0.45	-383.0310
6	CA-2014-115812	2014-06-09	2014-06-14	Standard Class	BH-11710	Brosina Hoffman	Consumer	United States	Los Angeles	California	90032	West	FUR-FU-10001487	Furniture	Furnishings	Eldon Expressions Wood and Plastic Desk Access...	48.8600	7	0.00	14.1694
11	CA-2014-115812	2014-06-09	2014-06-14	Standard Class	BH-11710	Brosina Hoffman	Consumer	United States	Los Angeles	California	90032	West	FUR-TA-10001539	Furniture	Tables	Chromcraft Rectangular Conference Tables	1706.1840	9	0.20	85.3092

In [167]:

data = data[['Order Date', 'Sales']]

In [168]:

data

Out[168]:

	Order Date	Sales
Row ID
1	2016-11-08	261.9600
2	2016-11-08	731.9400
4	2015-10-11	957.5775
6	2014-06-09	48.8600
11	2014-06-09	1706.1840
...	...	...
9963	2015-03-19	383.4656
9965	2016-12-05	13.4000
9981	2015-09-06	85.9800
9990	2014-01-21	25.2480
9991	2017-02-26	91.9600

2121 rows × 2 columns

In [169]:

data.sort_values('Order Date')  #Order Date 칼럼 기준으로 오름차순으로 불러오는 함수를 사용

Out[169]:

	Order Date	Sales
Row ID
7475	2014-01-06	2573.820
7661	2014-01-07	76.728
867	2014-01-10	51.940
717	2014-01-11	9.940
2979	2014-01-13	545.940
...	...	...
7637	2017-12-29	258.750
7633	2017-12-29	1207.840
5132	2017-12-29	101.120
7636	2017-12-29	300.980
907	2017-12-30	323.136

2121 rows × 2 columns

우리가 알고자하는 것은 시간에 따른 Sales의 트렌드 임으로 다시 정리를 해준다.

In [185]:

data = data.groupby('Order Date').sum()  # 각 날짜별로 총합을 구한다.

In [186]:

data

Out[186]:

	Sales
Order Date
2014-01-06	2573.8200
2014-01-07	76.7280
2014-01-10	51.9400
2014-01-11	9.9400
2014-01-13	879.9390
...	...
2017-12-24	1393.4940
2017-12-25	832.4540
2017-12-28	551.2568
2017-12-29	2330.7180
2017-12-30	323.1360

889 rows × 1 columns

In [187]:

data.index

Out[187]:

DatetimeIndex(['2014-01-06', '2014-01-07', '2014-01-10', '2014-01-11',
               '2014-01-13', '2014-01-14', '2014-01-16', '2014-01-19',
               '2014-01-20', '2014-01-21',
               ...
               '2017-12-18', '2017-12-19', '2017-12-21', '2017-12-22',
               '2017-12-23', '2017-12-24', '2017-12-25', '2017-12-28',
               '2017-12-29', '2017-12-30'],
              dtype='datetime64[ns]', name='Order Date', length=889, freq=None)

plot을 그려보자.

In [188]:

plt.figure(figsize = (20,10))
sns.lineplot(x = data.index, y = data['Sales'])

Out[188]:

<AxesSubplot:xlabel='Order Date', ylabel='Sales'>

위와 같이 컬럼이 한가지만 존재할 경우 그리고, x축에 들어갈 값이 (날짜 등과같은) 존재할 경우 아래와 같은 .plot로 더 간단하게 그릴 수 있다. (Seaborn 보다 plt가 편하다.)

In [189]:

data.plot(figsize = (20,10))

Out[189]:

<AxesSubplot:xlabel='Order Date'>

추세를 예측하기 어려운 그림이다. 월 단위로 묶어서 다시 보자. .resample() 함수롤 사용한다.

In [193]:

data = data.resample('MS').mean()  # MS는 월별, 그리고 평균값 mean() 

In [194]:

data  # 답을 비교하기위한 data

Out[194]:

	Sales
Order Date
2014-01-01	480.194231
2014-02-01	367.931600
2014-03-01	857.291529
2014-04-01	567.488357
2014-05-01	432.049187
2014-06-01	695.059242
2014-07-01	601.169500
2014-08-01	457.521656
2014-09-01	992.353367
2014-10-01	769.015437
2014-11-01	980.221486
2014-12-01	1532.298325
2015-01-01	978.328467
2015-02-01	522.395667
2015-03-01	781.236437
2015-04-01	805.822962
2015-05-01	624.996700
2015-06-01	428.565500
2015-07-01	719.706316
2015-08-01	602.412012
2015-09-01	1382.790684
2015-10-01	632.980184
2015-11-01	1286.701354
2015-12-01	1049.355418
2016-01-01	508.182867
2016-02-01	356.868273
2016-03-01	609.575810
2016-04-01	695.373158
2016-05-01	687.265227
2016-06-01	816.910750
2016-07-01	768.736412
2016-08-01	734.307782
2016-09-01	1135.953371
2016-10-01	624.872474
2016-11-01	1271.345152
2016-12-01	1410.719808
2017-01-01	397.602133
2017-02-01	528.179800
2017-03-01	544.672240
2017-04-01	453.297905
2017-05-01	678.302328
2017-06-01	826.460291
2017-07-01	562.524857
2017-08-01	857.881889
2017-09-01	1209.508583
2017-10-01	875.362728
2017-11-01	1277.817759
2017-12-01	1256.298672

In [195]:

new_df = data[:'2016']  # train data로 사용할 것이다.

In [196]:

from statsmodels.tsa.ar_model import AutoReg

In [197]:

model = AutoReg(new_df, lags = 12)

/home/ubuntu/.local/lib/python3.6/site-packages/statsmodels/tsa/ar_model.py:252: FutureWarning:

The parameter names will change after 0.12 is released. Set old_names to False to use the new names now. Set old_names to True to use the old names. 

In [198]:

model_fit = model.fit()

In [199]:

len(new_df)

Out[199]:

36

In [202]:

pred = model_fit.predict(start = len(new_df) , end = len(new_df)+11 )

In [203]:

pred

Out[203]:

2017-01-01     660.758250
2017-02-01     310.973369
2017-03-01     433.467903
2017-04-01     589.733771
2017-05-01     634.609408
2017-06-01     932.697417
2017-07-01     959.741129
2017-08-01     983.674137
2017-09-01     989.383080
2017-10-01     554.888314
2017-11-01    1030.961310
2017-12-01    1194.389892
Freq: MS, dtype: float64

In [204]:

ax = data.plot()
pred.plot(ax=ax, figsize=(20,10))

Out[204]:

<AxesSubplot:xlabel='Order Date'>

어느 정도 트렌드는 따라가지만, prophet보다 떨어진 성능을 확인할 수 있다.

• prophet 를 사용하는 것이 더 좋다.

• 출처 : 파이썬을 활용한 이커머스 데이터 분석