Chapter06.Random Forest

파이썬을 활용한 이커머스 데이터분석_강의를 듣고 따라했던 코딩과 요점을 정리하였다.

출처: fast campus

Chapter.06 프로모션 효율 예측 (Random Forest)¶

분석의 목적¶

Random Forest 를 이용하여, 프로모션에 반응할 고객을 예측
고객 데이터와 거래 데이터를 통합 활용

In [1]:

import pandas as pd
import numpy as np

import matplotlib.pyplot as plt
import seaborn as sns

In [2]:

mem = pd.read_csv('./data/member.csv')
tran = pd.read_csv('./data/transaction.csv')

In [3]:

mem.head(5)

Out[3]:

	id	recency	zip_code	is_referral	channel
0	906145	10	Surburban	0	Phone
1	184478	6	Rural	1	Web
2	394235	7	Surburban	1	Web
3	130152	9	Rural	1	Web
4	940352	2	Urban	0	Web

데이터 컬럼 살펴보기¶

id : 아이디 (의미가 없는 컬럼으로 추후 드랍예정이다.)
recency : 최근이용을 언제 했었나 (10일전, 6일전 ..)
zip_code : 우편번호 (한번 가공이 된 상태)
is_referral : 추천인이 있고 없음 으로 가입을 했는지
channe : 서비스 이용 채널 (폰, 웹 )
conversion : 프로모션을 받고 나서 고객이 구입을 했는지 안했는지 (우리가 예측하고자 하는 컬럼/종속변수)

In [4]:

mem.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 64000 entries, 0 to 63999
Data columns (total 6 columns):
 #   Column       Non-Null Count  Dtype 
---  ------       --------------  ----- 
 0   id           64000 non-null  int64 
 1   recency      64000 non-null  int64 
 2   zip_code     64000 non-null  object
 3   is_referral  64000 non-null  int64 
 4   channel      64000 non-null  object
 5   conversion   64000 non-null  int64 
dtypes: int64(4), object(2)
memory usage: 2.9+ MB

In [5]:

mem.describe()

Out[5]:

	id	recency	is_referral	conversion
count	64000.000000	64000.000000	64000.000000	64000.000000
mean	550694.137797	5.763734	0.502250	0.146781
std	259105.689773	3.507592	0.499999	0.353890
min	100001.000000	1.000000	0.000000	0.000000
25%	326772.000000	2.000000	0.000000	0.000000
50%	551300.000000	6.000000	1.000000	0.000000
75%	774914.500000	9.000000	1.000000	0.000000
max	999997.000000	12.000000	1.000000	1.000000

In [6]:

tran.head()

Out[6]:

	id	num_item	total_amount
0	906145	5	34000
1	906145	1	27000
2	906145	4	33000
3	184478	4	29000
4	394235	4	33000

데이터 컬럼 살펴보기¶

id : 아이디 (의미가 없는 컬럼으로 추후 드랍예정이다.)
num_item : 한 거래에 몇개에 아이템을 구매 했는지
total_amount : 총 금액

In [7]:

tran.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 196836 entries, 0 to 196835
Data columns (total 3 columns):
 #   Column        Non-Null Count   Dtype
---  ------        --------------   -----
 0   id            196836 non-null  int64
 1   num_item      196836 non-null  int64
 2   total_amount  196836 non-null  int64
dtypes: int64(3)
memory usage: 4.5 MB

In [8]:

tran.describe()

Out[8]:

	id	num_item	total_amount
count	196836.000000	196836.000000	196836.000000
mean	550557.552932	3.078365	21837.102969
std	259254.795613	1.478408	8218.005565
min	100001.000000	1.000000	8000.000000
25%	326719.000000	2.000000	15000.000000
50%	550918.000000	3.000000	22000.000000
75%	774916.000000	4.000000	29000.000000
max	999997.000000	6.000000	38000.000000

퓨처 엔지니어링을 해보자

In [9]:

tran

Out[9]:

	id	num_item	total_amount
0	906145	5	34000
1	906145	1	27000
2	906145	4	33000
3	184478	4	29000
4	394235	4	33000
...	...	...	...
196831	536246	5	24000
196832	927617	5	26000
196833	927617	3	22000
196834	927617	3	18000
196835	927617	3	20000

196836 rows × 3 columns

In [10]:

tran['total_amount'] / tran['num_item']

Out[10]:

0          6800.000000
1         27000.000000
2          8250.000000
3          7250.000000
4          8250.000000
              ...     
196831     4800.000000
196832     5200.000000
196833     7333.333333
196834     6000.000000
196835     6666.666667
Length: 196836, dtype: float64

평균 금액 컬럼을 하나더 만들고 싶어서, total_amount을 num_item로 나누는 결과값을 시리즈로 나온 것을 확인했다.

이것을 다시 기존 tran 데이터에 컬럼으로 추가해 보자. avg_prie 이란 컬럼으로

In [11]:

tran['avg_price'] = tran['total_amount'] / tran['num_item']

In [12]:

tran

Out[12]:

	id	num_item	total_amount	avg_price
0	906145	5	34000	6800.000000
1	906145	1	27000	27000.000000
2	906145	4	33000	8250.000000
3	184478	4	29000	7250.000000
4	394235	4	33000	8250.000000
...	...	...	...	...
196831	536246	5	24000	4800.000000
196832	927617	5	26000	5200.000000
196833	927617	3	22000	7333.333333
196834	927617	3	18000	6000.000000
196835	927617	3	20000	6666.666667

196836 rows × 4 columns

.groupby() 함수로 id 별로 인덱스해서 평균내어 데이터를 보자.

In [13]:

tran.groupby('id').mean()

Out[13]:

	num_item	total_amount	avg_price
id
100001	3.500000	26000.000000	7500.000000
100008	5.000000	26000.000000	5200.000000
100032	2.666667	20666.666667	9366.666667
100036	3.000000	25800.000000	13273.333333
100070	3.250000	21250.000000	8537.500000
...	...	...	...
999932	5.000000	32000.000000	6400.000000
999981	2.000000	22750.000000	12875.000000
999990	3.000000	28000.000000	10388.888889
999995	2.000000	27000.000000	13500.000000
999997	2.000000	13000.000000	6500.000000

64000 rows × 3 columns

In [14]:

tran['id'].value_counts()

Out[14]:

446874    5
473857    5
384266    5
648461    5
130318    5
         ..
674652    1
670546    1
192229    1
720615    1
789077    1
Name: id, Length: 64000, dtype: int64

tran.groupby('id').mean() 와 tran['id'].value_counts() 의 결과를 합쳐주자.

In [15]:

tran_mean = tran.groupby('id').mean()

In [16]:

tran_cnt = tran['id'].value_counts()

In [17]:

pd.concat([tran_mean, tran_cnt], axis=1)

Out[17]:

	num_item	total_amount	avg_price	id
100001	3.500000	26000.000000	7500.000000	2
100008	5.000000	26000.000000	5200.000000	1
100032	2.666667	20666.666667	9366.666667	3
100036	3.000000	25800.000000	13273.333333	5
100070	3.250000	21250.000000	8537.500000	4
...	...	...	...	...
999932	5.000000	32000.000000	6400.000000	1
999981	2.000000	22750.000000	12875.000000	4
999990	3.000000	28000.000000	10388.888889	3
999995	2.000000	27000.000000	13500.000000	1
999997	2.000000	13000.000000	6500.000000	1

64000 rows × 4 columns

In [18]:

tran_df = pd.concat([tran_mean, tran_cnt], axis=1)

id 의 컬럼을 이름을 바꿔주자.

In [19]:

tran_df.rename(columns = {'id':'count'})

Out[19]:

	num_item	total_amount	avg_price	count
100001	3.500000	26000.000000	7500.000000	2
100008	5.000000	26000.000000	5200.000000	1
100032	2.666667	20666.666667	9366.666667	3
100036	3.000000	25800.000000	13273.333333	5
100070	3.250000	21250.000000	8537.500000	4
...	...	...	...	...
999932	5.000000	32000.000000	6400.000000	1
999981	2.000000	22750.000000	12875.000000	4
999990	3.000000	28000.000000	10388.888889	3
999995	2.000000	27000.000000	13500.000000	1
999997	2.000000	13000.000000	6500.000000	1

64000 rows × 4 columns

이제는 tran 과 mem 데이터를 하나로 합칠 것이다.

In [20]:

mem

Out[20]:

	id	recency	zip_code	is_referral	channel	conversion
0	906145	10	Surburban	0	Phone	0
1	184478	6	Rural	1	Web	0
2	394235	7	Surburban	1	Web	0
3	130152	9	Rural	1	Web	0
4	940352	2	Urban	0	Web	0
...	...	...	...	...	...	...
63995	838295	10	Urban	0	Web	0
63996	547316	5	Urban	1	Phone	0
63997	131575	6	Urban	1	Phone	0
63998	603659	1	Surburban	1	Multichannel	0
63999	254229	1	Surburban	0	Web	0

64000 rows × 6 columns

In [21]:

mem.set_index('id', inplace=True)

In [22]:

mem

Out[22]:

	recency	zip_code	is_referral	channel	conversion
id
906145	10	Surburban	0	Phone	0
184478	6	Rural	1	Web	0
394235	7	Surburban	1	Web	0
130152	9	Rural	1	Web	0
940352	2	Urban	0	Web	0
...	...	...	...	...	...
838295	10	Urban	0	Web	0
547316	5	Urban	1	Phone	0
131575	6	Urban	1	Phone	0
603659	1	Surburban	1	Multichannel	0
254229	1	Surburban	0	Web	0

64000 rows × 5 columns

In [23]:

mem.join(tran_df)

Out[23]:

	recency	zip_code	is_referral	channel	conversion	num_item	total_amount	avg_price	id
id
906145	10	Surburban	0	Phone	0	3.333333	31333.333333	14016.666667	3
184478	6	Rural	1	Web	0	4.000000	29000.000000	7250.000000	1
394235	7	Surburban	1	Web	0	4.000000	20500.000000	5125.000000	2
130152	9	Rural	1	Web	0	1.750000	20750.000000	14875.000000	4
940352	2	Urban	0	Web	0	3.000000	31000.000000	10333.333333	1
...	...	...	...	...	...	...	...	...	...
838295	10	Urban	0	Web	0	3.500000	26000.000000	8012.500000	4
547316	5	Urban	1	Phone	0	1.800000	17800.000000	11300.000000	5
131575	6	Urban	1	Phone	0	4.000000	30500.000000	7833.333333	2
603659	1	Surburban	1	Multichannel	0	3.200000	21600.000000	7583.333333	5
254229	1	Surburban	0	Web	0	3.400000	24400.000000	9280.000000	5

64000 rows × 9 columns

2 데이터를 합친 것을 하나의 data 로 할당해주자

In [26]:

data = mem.join(tran_df)

결측치 확인

In [28]:

data.isna().sum()

Out[28]:

recency         0
zip_code        0
is_referral     0
channel         0
conversion      0
num_item        0
total_amount    0
avg_price       0
id              0
dtype: int64

In [30]:

data.isna().sum() / len(data)

Out[30]:

recency         0.0
zip_code        0.0
is_referral     0.0
channel         0.0
conversion      0.0
num_item        0.0
total_amount    0.0
avg_price       0.0
id              0.0
dtype: float64

카테고리 벨류, 즉 텍스트 값 처리하기

우선 유니크값 벨류 확인하기

In [37]:

data['zip_code'].nunique()

Out[37]:

zip_code에는 3개의 벨류 , channel 에도 3개의 벨류 확인

In [38]:

data['zip_code'].unique()

Out[38]:

array(['Surburban', 'Rural', 'Urban'], dtype=object)

In [39]:

data['channel'].nunique()

Out[39]:

In [40]:

data['channel'].unique()

Out[40]:

array(['Phone', 'Web', 'Multichannel'], dtype=object)

.get_dummies() 로 텍스트 벨류 처리하기

In [44]:

pd.get_dummies(data, columns=['zip_code', 'channel'], drop_first = True)

Out[44]:

	recency	is_referral	conversion	num_item	total_amount	avg_price	id	zip_code_Surburban	zip_code_Urban	channel_Phone	channel_Web
id
906145	10	0	0	3.333333	31333.333333	14016.666667	3	1	0	1	0
184478	6	1	0	4.000000	29000.000000	7250.000000	1	0	0	0	1
394235	7	1	0	4.000000	20500.000000	5125.000000	2	1	0	0	1
130152	9	1	0	1.750000	20750.000000	14875.000000	4	0	0	0	1
940352	2	0	0	3.000000	31000.000000	10333.333333	1	0	1	0	1
...	...	...	...	...	...	...	...	...	...	...	...
838295	10	0	0	3.500000	26000.000000	8012.500000	4	0	1	0	1
547316	5	1	0	1.800000	17800.000000	11300.000000	5	0	1	1	0
131575	6	1	0	4.000000	30500.000000	7833.333333	2	0	1	1	0
603659	1	1	0	3.200000	21600.000000	7583.333333	5	1	0	0	0
254229	1	0	0	3.400000	24400.000000	9280.000000	5	1	0	0	1

64000 rows × 11 columns

In [45]:

data = pd.get_dummies(data, columns=['zip_code', 'channel'], drop_first = True)

이것으로 data 전처리는 끝났고 준비가 되었다. 모델링을 해보자.

In [47]:

from sklearn.model_selection import train_test_split

In [56]:

X = data.drop('conversion', axis = 1)
y = data['conversion']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.3, random_state = 100 )

In [57]:

from sklearn.ensemble import RandomForestClassifier

In [58]:

model = RandomForestClassifier(max_depth = 10, random_state = 100)

In [59]:

model.fit(X_train, y_train)

Out[59]:

RandomForestClassifier(max_depth=10, random_state=100)

훈련이 끝났으니, 예측을 해보자

In [61]:

pred = model.predict(X_test)

In [62]:

y_test

Out[62]:

id
632233    0
412308    0
184792    0
546903    0
113517    0
         ..
629047    0
470260    0
673575    0
345057    0
182726    0
Name: conversion, Length: 19200, dtype: int64

평가도 해보자.

In [64]:

from sklearn.metrics import accuracy_score, confusion_matrix

In [65]:

accuracy_score(y_test, pred)

Out[65]:

0.87515625

정확도는 87%로 나쁘지 않게나왔다. 무조건 좋은 치수도 아니다.

In [66]:

confusion_matrix(y_test, pred)

Out[66]:

array([[16403,    60],
       [ 2337,   400]])

0 이라고 예상하고 실제가 0 인 수치가: 16403, 1 이라고 예상하고 실제로도 1이 였던 것이 400건, 예상이 0이라고 하고 실제로는 1이 였던 것은 2337건 , 예상은 1이고 실제는 0 이 었던 수치는 60건 이다.

전체적인 평가를 해보았을때, 0 이라고 예상했는데 실제는 1이 였던것이 2337이면 정확도가 그렇게 높지 않은 것으로 판단 할 수 있다.

그래서 다른 방법으로 정확도를 높여 보자.

In [67]:

from sklearn.metrics import classification_report

In [69]:

classification_report(y_test, pred)

Out[69]:

'              precision    recall  f1-score   support\n\n           0       0.88      1.00      0.93     16463\n           1       0.87      0.15      0.25      2737\n\n    accuracy                           0.88     19200\n   macro avg       0.87      0.57      0.59     19200\nweighted avg       0.87      0.88      0.83     19200\n'

이쁘게 나오지 않는다. 앞에 print를 넣어 출력하자.

In [70]:

print(classification_report(y_test, pred))

              precision    recall  f1-score   support

           0       0.88      1.00      0.93     16463
           1       0.87      0.15      0.25      2737

    accuracy                           0.88     19200
   macro avg       0.87      0.57      0.59     19200
weighted avg       0.87      0.88      0.83     19200

1 에대한 recall 값과 f1-score 값이 상대적으로 너무 낮게 나왔다.

Regressor 로 평가해보기.

In [73]:

from sklearn.ensemble import RandomForestRegressor

In [74]:

rf = RandomForestRegressor(max_depth = 10, random_state = 100 )

In [75]:

rf.fit(X_train, y_train)

Out[75]:

RandomForestRegressor(max_depth=10, random_state=100)

학습이 된 것으로 예측

In [76]:

rf.predict(X_test)

Out[76]:

array([0.        , 0.05858361, 0.        , ..., 0.11632305, 0.        ,
       0.05458311])

In [77]:

pred = rf.predict(X_test)

In [78]:

pd.DataFrame(pred)

Out[78]:

	0
0	0.000000
1	0.058584
2	0.000000
3	0.594169
4	0.083316
...	...
19195	0.142431
19196	0.000000
19197	0.116323
19198	0.000000
19199	0.054583

19200 rows × 1 columns

In [79]:

y_test

Out[79]:

id
632233    0
412308    0
184792    0
546903    0
113517    0
         ..
629047    0
470260    0
673575    0
345057    0
182726    0
Name: conversion, Length: 19200, dtype: int64

비교를 하기 위해서 pred의 값이 0 아니면 1로 나와야 한다. 그래서 다음의 추가 작업을 진행해보자.

In [80]:

def conv(x):
    if x >= 0.5:
        return 1
    else:
        return 0
    

In [91]:

pd_result =pd.Series(pred).apply(lambda x: conv(x))

In [92]:

pd_result

Out[92]:

0        0
1        0
2        0
3        1
4        0
        ..
19195    0
19196    0
19197    0
19198    0
19199    0
Length: 19200, dtype: int64

또 다른 방법으로는

In [89]:

result = []
for i in pred:
    result.append(conv(i))

In [90]:

result

Out[90]:

[0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 1,
 1,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 1,
 1,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 1,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 ...]

또 다른 방법으로는

In [94]:

result_comp = [1 if x >= 0.5 else 0 for x in pred]

In [95]:

result_comp

Out[95]:

[0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 1,
 1,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 1,
 1,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 1,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 0,
 1,
 ...]

이제 비교를 해보자.

In [96]:

accuracy_score(y_test, result_comp)

Out[96]:

0.8799479166666667

In [97]:

confusion_matrix(y_test, result_comp)

Out[97]:

array([[16313,   150],
       [ 2155,   582]])

In [98]:

print(classification_report(y_test, result_comp))

              precision    recall  f1-score   support

           0       0.88      0.99      0.93     16463
           1       0.80      0.21      0.34      2737

    accuracy                           0.88     19200
   macro avg       0.84      0.60      0.63     19200
weighted avg       0.87      0.88      0.85     19200

0.5 가 아닌 0.3 으로 정해서, 조절해보자.

In [99]:

result_3 = [1 if x >= 0.3 else 0 for x in pred]

In [100]:

print(classification_report(y_test, result_3))

              precision    recall  f1-score   support

           0       0.90      0.96      0.92     16463
           1       0.55      0.33      0.42      2737

    accuracy                           0.87     19200
   macro avg       0.72      0.64      0.67     19200
weighted avg       0.85      0.87      0.85     19200

비교를 한번해보자 . max_depth 값을 조정해서.

In [103]:

rf = RandomForestRegressor(max_depth = 10, random_state = 100 )
rf.fit(X_train, y_train)
pred = rf.predict(X_test)
result_comp = [1 if x >= 0.5 else 0 for x in pred]

accuracy_score(y_test, result_comp)

Out[103]:

0.8799479166666667

In [104]:

rf = RandomForestRegressor(max_depth = 12, random_state = 100 )
rf.fit(X_train, y_train)
pred = rf.predict(X_test)
result_comp = [1 if x >= 0.5 else 0 for x in pred]

accuracy_score(y_test, result_comp)

Out[104]:

0.8805729166666667

max_depth = 12로 했을때, 조금더 정확도가 높아졌다. 이번엔 n_estimators = 150 으로 조정해보자

In [106]:

rf = RandomForestRegressor(n_estimators = 150, max_depth = 12, random_state = 100 )
rf.fit(X_train, y_train)
pred = rf.predict(X_test)
result_comp = [1 if x >= 0.5 else 0 for x in pred]

accuracy_score(y_test, result_comp)

Out[106]:

0.8800520833333333

결과는 그냥 기본 값이 더 나은 결과를 보여준다. n_estimators = 100 (기본값은 100이다)

이번엔 min_samples_leaf 값을 조정해보자.

In [108]:

rf = RandomForestRegressor(n_estimators = 100, max_depth = 12, random_state = 100 ,min_samples_leaf = 5)
rf.fit(X_train, y_train)
pred = rf.predict(X_test)
result_comp = [1 if x >= 0.5 else 0 for x in pred]

accuracy_score(y_test, result_comp)

Out[108]:

0.8810416666666666

조금 더 좋은 값이 나왔다. 하지만 지금의 조건 준것이 가장 best 는 아님을 알아야한다. 여러개의 파라미터 조정을 통해 이것은 언제든 변할 수 있다.

각각의 변수 중요성을 살펴보자.

In [110]:

rf.feature_importances_

Out[110]:

array([0.06390282, 0.0229547 , 0.31878145, 0.15857333, 0.25092914,
       0.14468844, 0.00936383, 0.00914032, 0.01179997, 0.009866  ])

In [111]:

X_train.columns

Out[111]:

Index(['recency', 'is_referral', 'num_item', 'total_amount', 'avg_price', 'id',
       'zip_code_Surburban', 'zip_code_Urban', 'channel_Phone', 'channel_Web'],
      dtype='object')

그래프로 만들어 보자.

In [114]:

plt.figure(figsize= (20, 10))
sns.barplot(x = X_train.columns, y = rf.feature_importances_)

Out[114]:

<AxesSubplot:>

출처 :fast campus_파이썬을 활용한 이커머스 데이터 분석

'파이썬을 활용한 이커머스 데이터 분석' 카테고리의 다른 글

Chapter08.Times Series 쇼핑몰 매출 예측 (Times Series) (0)	2021.07.08
Chapter.07 고객 분류 (Kmeans) (0)	2021.06.14
Chapter05.구매 요인 분석(Dicision Tree) (0)	2021.06.12
Chapter04.KNN (0)	2021.06.10
Chapter03. 광고 반응률 예측 (Logistic Regression) (0)	2021.06.10

미래를 위한 취미

Chapter.06 프로모션 효율 예측 (Random Forest)

Chapter.06 프로모션 효율 예측 (Random Forest)¶

분석의 목적¶

데이터 컬럼 살펴보기¶

데이터 컬럼 살펴보기¶

'파이썬을 활용한 이커머스 데이터 분석' 카테고리의 다른 글

티스토리툴바

Chapter.06 프로모션 효율 예측 (Random Forest)

Chapter.06 프로모션 효율 예측 (Random Forest)¶

분석의 목적¶

데이터 컬럼 살펴보기¶

데이터 컬럼 살펴보기¶

'파이썬을 활용한 이커머스 데이터 분석' 카테고리의 다른 글

'파이썬을 활용한 이커머스 데이터 분석' Related Articles

티스토리툴바