E-Commerce Part Ⅶ: 시계열 분석 응용

Time Series Data Analysis¶

 

 

ⅰ. Importing Modules & Data Skimming¶

In [1]:

import numpy as np
import pandas as pd

import matplotlib.pyplot as plt
import seaborn as sns

In [2]:

pd.options.display.max_columns=30

df = pd.read_excel("Data/Superstore.xls", index_col=0)
df.head()

Out[2]:

	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 [3]:

df.describe()

Out[3]:

	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

 

 

ⅱ. Data Filtering¶

In [4]:

df['Category'].value_counts()

Out[4]:

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

In [5]:

df_furniture =df[df['Category'] == 'Furniture']
df_furniture.head()

Out[5]:

	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 [6]:

df_furniture = df_furniture[['Order Date', 'Sales']].sort_values('Order Date')
df_furniture

Out[6]:

	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

In [7]:

df_furniture = df_furniture.groupby('Order Date').sum()
df_furniture

Out[7]:

	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 [8]:

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

Out[8]:

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

 

일 단위로는 매출이 너무 들쭉날쭉해서 의미있는 시계열 데이터 분석이 어렵다 → 월 단위로 데이터 묶어본 다음 다시 시각화

 

In [9]:

# 월별 평균 매출액
df_furniture_month = df_furniture.resample('MS').mean()
df_furniture_month.head()

Out[9]:

	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	432.049188

In [10]:

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

Out[10]:

<AxesSubplot:xlabel='Order Date'>

 

 

ⅲ. Modeling by FBProphet Lib¶

In [11]:

from fbprophet import Prophet

# prophet을 사용하기 위한 DataFrame 사전작업
# 시계열변수 col명을 ds 로 설정
# 종속변수 col명을 y 로 설정

df = df_furniture_month.reset_index()
df.columns = ['ds', 'y']
df.head()

Out[11]:

	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	432.049188

In [12]:

model = Prophet()
model.fit(df)

INFO:numexpr.utils:NumExpr defaulting to 8 threads.
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[12]:

<fbprophet.forecaster.Prophet at 0x25a22439df0>

In [13]:

# 시계열 데이터를 예측할 범위 데이터 프레임
future = model.make_future_dataframe(periods = 12, freq = 'MS')

pred = model.predict(future)
pred.head()

Out[13]:

	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	378.598521	690.510121	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	290.790180	599.278408	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	544.490813	854.041798	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	474.123664	786.956816	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	405.451940	719.629716	731.617622	731.617622	-160.815662	-160.815662	-160.815662	-160.815662	-160.815662	-160.815662	0.0	0.0	0.0	570.801960

In [14]:

pred = pred[['ds', 'yhat', 'yhat_lower', 'yhat_upper']]
pred.head()

Out[14]:

	ds	yhat	yhat_lower	yhat_upper
0	2014-01-01	535.372051	378.598521	690.510121
1	2014-02-01	451.116320	290.790180	599.278408
2	2014-03-01	706.401580	544.490813	854.041798
3	2014-04-01	630.086487	474.123664	786.956816
4	2014-05-01	570.801960	405.451940	719.629716

In [15]:

model.plot(pred)
plt.show()

 

 

ⅳ. Validating the Predicted Values from Model¶

In [16]:

df_new = df_furniture_month[:'2016']
df_new.tail()

Out[16]:

	Sales
Order Date
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

In [17]:

df_new = df_new.reset_index()
df_new.columns = ['ds', 'y']
df_new.head()

Out[17]:

	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	432.049188

In [18]:

model = Prophet()
model.fit(df_new)

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[18]:

<fbprophet.forecaster.Prophet at 0x25a2462da60>

In [19]:

future = model.make_future_dataframe(periods = 12, freq = 'MS')
pred = model.predict(future)

pred = pred[['ds', 'yhat']].set_index('ds')

pred = pred['2017':]
pred

Out[19]:

	yhat
ds
2017-01-01	804.654438
2017-02-01	601.206518
2017-03-01	782.905069
2017-04-01	754.585297
2017-05-01	685.352803
2017-06-01	805.507654
2017-07-01	792.293714
2017-08-01	714.448420
2017-09-01	1208.615981
2017-10-01	747.521868
2017-11-01	1271.011086
2017-12-01	1473.348777

In [20]:

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

Out[20]:

<AxesSubplot:xlabel='ds'>

Prophet 모델을 통해 얻은 결과값과 실제값을 비교한 plot

 

In [21]:

model = Prophet()
model.fit(df_new)

future = model.make_future_dataframe(periods = 12, freq = 'MS')
pred = model.predict(future)

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.

In [22]:

model.plot_components(pred)
plt.show()