In [15]:
import pandas as pd
import pandas as pd
import seaborn as sns
import plotly.express as px
import plotly.graph_objects as plt
import matplotlib.pyplot as plt
%matplotlib inline
import warnings
warnings.filterwarnings('ignore')
In [16]:
# 加载数据
main_df = pd.read_csv('marketing_campaign.csv', sep='\t')
df = main_df.copy()
df.head(10)
Out[16]:
ID Year_Birth Education Marital_Status Income Kidhome Teenhome Dt_Customer Recency MntWines ... NumWebVisitsMonth AcceptedCmp3 AcceptedCmp4 AcceptedCmp5 AcceptedCmp1 AcceptedCmp2 Complain Z_CostContact Z_Revenue Response
0 5524 1957 Graduation Single 58138.0 0 0 04-09-2012 58 635 ... 7 0 0 0 0 0 0 3 11 1
1 2174 1954 Graduation Single 46344.0 1 1 08-03-2014 38 11 ... 5 0 0 0 0 0 0 3 11 0
2 4141 1965 Graduation Together 71613.0 0 0 21-08-2013 26 426 ... 4 0 0 0 0 0 0 3 11 0
3 6182 1984 Graduation Together 26646.0 1 0 10-02-2014 26 11 ... 6 0 0 0 0 0 0 3 11 0
4 5324 1981 PhD Married 58293.0 1 0 19-01-2014 94 173 ... 5 0 0 0 0 0 0 3 11 0
5 7446 1967 Master Together 62513.0 0 1 09-09-2013 16 520 ... 6 0 0 0 0 0 0 3 11 0
6 965 1971 Graduation Divorced 55635.0 0 1 13-11-2012 34 235 ... 6 0 0 0 0 0 0 3 11 0
7 6177 1985 PhD Married 33454.0 1 0 08-05-2013 32 76 ... 8 0 0 0 0 0 0 3 11 0
8 4855 1974 PhD Together 30351.0 1 0 06-06-2013 19 14 ... 9 0 0 0 0 0 0 3 11 1
9 5899 1950 PhD Together 5648.0 1 1 13-03-2014 68 28 ... 20 1 0 0 0 0 0 3 11 0

10 rows × 29 columns

数据分析¶

In [17]:
#shape of the dataset
df.shape
Out[17]:
(2240, 29)
In [18]:
# basic information of dataset
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 2240 entries, 0 to 2239
Data columns (total 29 columns):
 #   Column               Non-Null Count  Dtype  
---  ------               --------------  -----  
 0   ID                   2240 non-null   int64  
 1   Year_Birth           2240 non-null   int64  
 2   Education            2240 non-null   object 
 3   Marital_Status       2240 non-null   object 
 4   Income               2216 non-null   float64
 5   Kidhome              2240 non-null   int64  
 6   Teenhome             2240 non-null   int64  
 7   Dt_Customer          2240 non-null   object 
 8   Recency              2240 non-null   int64  
 9   MntWines             2240 non-null   int64  
 10  MntFruits            2240 non-null   int64  
 11  MntMeatProducts      2240 non-null   int64  
 12  MntFishProducts      2240 non-null   int64  
 13  MntSweetProducts     2240 non-null   int64  
 14  MntGoldProds         2240 non-null   int64  
 15  NumDealsPurchases    2240 non-null   int64  
 16  NumWebPurchases      2240 non-null   int64  
 17  NumCatalogPurchases  2240 non-null   int64  
 18  NumStorePurchases    2240 non-null   int64  
 19  NumWebVisitsMonth    2240 non-null   int64  
 20  AcceptedCmp3         2240 non-null   int64  
 21  AcceptedCmp4         2240 non-null   int64  
 22  AcceptedCmp5         2240 non-null   int64  
 23  AcceptedCmp1         2240 non-null   int64  
 24  AcceptedCmp2         2240 non-null   int64  
 25  Complain             2240 non-null   int64  
 26  Z_CostContact        2240 non-null   int64  
 27  Z_Revenue            2240 non-null   int64  
 28  Response             2240 non-null   int64  
dtypes: float64(1), int64(25), object(3)
memory usage: 507.6+ KB
  • Here we have only 3 object type datatype and rest are numerical
In [19]:
# 检查特征唯一值数量
df.nunique()
Out[19]:
ID                     2240
Year_Birth               59
Education                 5
Marital_Status            8
Income                 1974
Kidhome                   3
Teenhome                  3
Dt_Customer             663
Recency                 100
MntWines                776
MntFruits               158
MntMeatProducts         558
MntFishProducts         182
MntSweetProducts        177
MntGoldProds            213
NumDealsPurchases        15
NumWebPurchases          15
NumCatalogPurchases      14
NumStorePurchases        14
NumWebVisitsMonth        16
AcceptedCmp3              2
AcceptedCmp4              2
AcceptedCmp5              2
AcceptedCmp1              2
AcceptedCmp2              2
Complain                  2
Z_CostContact             1
Z_Revenue                 1
Response                  2
dtype: int64
  • 在上面的单元格中，“Z_CostContact”和“Z_Revenue”在所有行中都有相同的值，这就是为什么它们不会在模型构建中做出任何贡献。 这样我们就可以放弃它们
In [20]:
# 检查空值
df.isna().any()
Out[20]:
ID                     False
Year_Birth             False
Education              False
Marital_Status         False
Income                  True
Kidhome                False
Teenhome               False
Dt_Customer            False
Recency                False
MntWines               False
MntFruits              False
MntMeatProducts        False
MntFishProducts        False
MntSweetProducts       False
MntGoldProds           False
NumDealsPurchases      False
NumWebPurchases        False
NumCatalogPurchases    False
NumStorePurchases      False
NumWebVisitsMonth      False
AcceptedCmp3           False
AcceptedCmp4           False
AcceptedCmp5           False
AcceptedCmp1           False
AcceptedCmp2           False
Complain               False
Z_CostContact          False
Z_Revenue              False
Response               False
dtype: bool
In [21]:
# Checking number of null values
df.isnull().sum()
Out[21]:
ID                      0
Year_Birth              0
Education               0
Marital_Status          0
Income                 24
Kidhome                 0
Teenhome                0
Dt_Customer             0
Recency                 0
MntWines                0
MntFruits               0
MntMeatProducts         0
MntFishProducts         0
MntSweetProducts        0
MntGoldProds            0
NumDealsPurchases       0
NumWebPurchases         0
NumCatalogPurchases     0
NumStorePurchases       0
NumWebVisitsMonth       0
AcceptedCmp3            0
AcceptedCmp4            0
AcceptedCmp5            0
AcceptedCmp1            0
AcceptedCmp2            0
Complain                0
Z_CostContact           0
Z_Revenue               0
Response                0
dtype: int64
  • 收入列中有一些缺失值，因此我们需要用maean或中位数来填充它。
In [22]:
# 使用热力图检查空值
sns.heatmap(df.isnull())
Out[22]:
<Axes: >
In [23]:
# 删除该列，因为它们不会对模型构建做出贡献
df = df.drop(columns=["Z_CostContact", "Z_Revenue"], axis=1)
df.head(10)
Out[23]:
ID Year_Birth Education Marital_Status Income Kidhome Teenhome Dt_Customer Recency MntWines ... NumCatalogPurchases NumStorePurchases NumWebVisitsMonth AcceptedCmp3 AcceptedCmp4 AcceptedCmp5 AcceptedCmp1 AcceptedCmp2 Complain Response
0 5524 1957 Graduation Single 58138.0 0 0 04-09-2012 58 635 ... 10 4 7 0 0 0 0 0 0 1
1 2174 1954 Graduation Single 46344.0 1 1 08-03-2014 38 11 ... 1 2 5 0 0 0 0 0 0 0
2 4141 1965 Graduation Together 71613.0 0 0 21-08-2013 26 426 ... 2 10 4 0 0 0 0 0 0 0
3 6182 1984 Graduation Together 26646.0 1 0 10-02-2014 26 11 ... 0 4 6 0 0 0 0 0 0 0
4 5324 1981 PhD Married 58293.0 1 0 19-01-2014 94 173 ... 3 6 5 0 0 0 0 0 0 0
5 7446 1967 Master Together 62513.0 0 1 09-09-2013 16 520 ... 4 10 6 0 0 0 0 0 0 0
6 965 1971 Graduation Divorced 55635.0 0 1 13-11-2012 34 235 ... 3 7 6 0 0 0 0 0 0 0
7 6177 1985 PhD Married 33454.0 1 0 08-05-2013 32 76 ... 0 4 8 0 0 0 0 0 0 0
8 4855 1974 PhD Together 30351.0 1 0 06-06-2013 19 14 ... 0 2 9 0 0 0 0 0 0 1
9 5899 1950 PhD Together 5648.0 1 1 13-03-2014 68 28 ... 0 0 20 1 0 0 0 0 0 0

10 rows × 27 columns

  • 让我们计算出过去两年客户投诉的数量以及正面或负面的回复数量。
In [24]:
# Complain: 1 if customer complained in the last 2 years, 0 otherwise
label_complain = ["No Complain","Complain"]

count_complain = pd.value_counts(df['Complain'], sort=True)
count_complain.plot(kind='bar', rot=0,color=['Green','Red'])
plt.title("Complain class Distribution")
plt.xticks(range(2),label_complain)
plt.xlabel("Complain")
plt.ylabel("Count of Complain")
Out[24]:
Text(0, 0.5, 'Count of Complain')
In [25]:
df['Complain'].value_counts()
# 1 if customer complained in the last 2 years, 0 otherwise
Out[25]:
0    2219
1      21
Name: Complain, dtype: int64
  • 从上图可以看出，顾客的投诉并不多。
In [26]:
# 让我们检查一下响应

# 响应：如果客户在最近 2 次活动中接受了优惠，则为 1，否则为 0
label_response = ["Denied","Accepted"]

count_response = pd.value_counts(df['Response'], sort=True)
count_response.plot(kind='bar', rot=0,color=['Green','Red'])
plt.title("Reponse class Distribution")
plt.xticks(range(2),label_response)
plt.xlabel("Respons")
plt.ylabel("Count of Response")
Out[26]:
Text(0, 0.5, 'Count of Response')
In [27]:
df['Response'].value_counts()
Out[27]:
0    1906
1     334
Name: Response, dtype: int64
  • 此图显示最后的优惠已被大多数客户拒绝

查看所有情况

  • AcceptedCmp1：如果客户在第一个活动中接受了优惠，则为 1，否则为 0
  • AcceptedCmp2：如果客户在第二次活动中接受了优惠，则为 1，否则为 0
  • AcceptedCmp3：如果客户在第三次活动中接受了优惠，则为 1，否则为 0
  • AcceptedCmp4：如果客户在第四次活动中接受了优惠，则为 1，否则为 0
  • AcceptedCmp5：如果客户在第五次活动中接受了优惠，则为 1，否则为 0
  • 响应：如果客户在上次活动中接受了优惠，则为 1，否则为 0
In [28]:
#Campagin 1

labels_c1 = ["Denied", "Accepted"]

count_c1 = pd.value_counts(df['AcceptedCmp1'], sort=True)
count_c1.plot(kind='bar', rot=0,color=['Green','Red'])
plt.title("AcceptedCmp1 class Distribution")
plt.xticks(range(2),labels_c1)
plt.xlabel("Campaign 1")
plt.ylabel("Count of Campagin1")
Out[28]:
Text(0, 0.5, 'Count of Campagin1')
In [29]:
df['AcceptedCmp1'].value_counts()
Out[29]:
0    2096
1     144
Name: AcceptedCmp1, dtype: int64
In [30]:
#Campagin 2

labels_c2 = ["Denied", "Accepted"]

count_c2 = pd.value_counts(df['AcceptedCmp2'], sort=True)
count_c2.plot(kind='bar', rot=0,color=['Green','Red'])
plt.title("AcceptedCmp2 class Distribution")
plt.xticks(range(2),labels_c2)
plt.xlabel("Campaign 2")
plt.ylabel("Count of Campagin2")
Out[30]:
Text(0, 0.5, 'Count of Campagin2')
In [31]:
df["AcceptedCmp2"].value_counts()
Out[31]:
0    2210
1      30
Name: AcceptedCmp2, dtype: int64
In [32]:
#Campagin 3

labels_c3 = ["Denied", "Accepted"]

count_c3 = pd.value_counts(df['AcceptedCmp3'], sort=True)
count_c3.plot(kind='bar', rot=0,color=['Green','Red'])
plt.title("AcceptedCmp3 class Distribution")
plt.xticks(range(2),labels_c3)
plt.xlabel("Campaign 3")
plt.ylabel("Count of Campagin3")
Out[32]:
Text(0, 0.5, 'Count of Campagin3')
In [33]:
df["AcceptedCmp3"].value_counts()
Out[33]:
0    2077
1     163
Name: AcceptedCmp3, dtype: int64
In [34]:
#Campagin 4

labels_c4 = ["Denied", "Accepted"]

count_c4 = pd.value_counts(df['AcceptedCmp4'], sort=True)
count_c4.plot(kind='bar', rot=0,color=['Green','Red'])
plt.title("AcceptedCmp4 class Distribution")
plt.xticks(range(2),labels_c4)
plt.xlabel("Campaign 4")
plt.ylabel("Count of Campagin4")
Out[34]:
Text(0, 0.5, 'Count of Campagin4')
In [35]:
df["AcceptedCmp4"].value_counts()
Out[35]:
0    2073
1     167
Name: AcceptedCmp4, dtype: int64

活动接受度比较。

  • 从上面的数据我们可以清楚地看到，所有活动中的大部分优惠都被顾客拒绝了。
  • 但活动 4 的接受度更高。
  • 活动 4 > 活动 3 > 活动 1 > 活动 2
In [36]:
#Finding the correlation between the feature column

plt.figure(figsize=(20,18))
sns.heatmap(df.corr(), annot=True)
plt.show()
  • 没有两列彼此有太多相关性，因此我们不能删除任何列

数据预处理¶

In [37]:
# Filling the missing value in the income by mean
df['Income'] = df['Income'].fillna(df['Income'].mean())
df.isnull().sum()
Out[37]:
ID                     0
Year_Birth             0
Education              0
Marital_Status         0
Income                 0
Kidhome                0
Teenhome               0
Dt_Customer            0
Recency                0
MntWines               0
MntFruits              0
MntMeatProducts        0
MntFishProducts        0
MntSweetProducts       0
MntGoldProds           0
NumDealsPurchases      0
NumWebPurchases        0
NumCatalogPurchases    0
NumStorePurchases      0
NumWebVisitsMonth      0
AcceptedCmp3           0
AcceptedCmp4           0
AcceptedCmp5           0
AcceptedCmp1           0
AcceptedCmp2           0
Complain               0
Response               0
dtype: int64
  • 数据集中没有空值
In [38]:
df.head()
Out[38]:
ID Year_Birth Education Marital_Status Income Kidhome Teenhome Dt_Customer Recency MntWines ... NumCatalogPurchases NumStorePurchases NumWebVisitsMonth AcceptedCmp3 AcceptedCmp4 AcceptedCmp5 AcceptedCmp1 AcceptedCmp2 Complain Response
0 5524 1957 Graduation Single 58138.0 0 0 04-09-2012 58 635 ... 10 4 7 0 0 0 0 0 0 1
1 2174 1954 Graduation Single 46344.0 1 1 08-03-2014 38 11 ... 1 2 5 0 0 0 0 0 0 0
2 4141 1965 Graduation Together 71613.0 0 0 21-08-2013 26 426 ... 2 10 4 0 0 0 0 0 0 0
3 6182 1984 Graduation Together 26646.0 1 0 10-02-2014 26 11 ... 0 4 6 0 0 0 0 0 0 0
4 5324 1981 PhD Married 58293.0 1 0 19-01-2014 94 173 ... 3 6 5 0 0 0 0 0 0 0

5 rows × 27 columns

In [39]:
# 检查“Marital_Status”中存在的唯一类别的数量

df['Marital_Status'].value_counts()
Out[39]:
Married     864
Together    580
Single      480
Divorced    232
Widow        77
Alone         3
Absurd        2
YOLO          2
Name: Marital_Status, dtype: int64
In [40]:
df['Marital_Status'] = df['Marital_Status'].replace(['Married', 'Together'],'relationship')
df['Marital_Status'] = df['Marital_Status'].replace(['Divorced', 'Widow', 'Alone', 'YOLO', 'Absurd','single'],'Single')
  • 在上面的单元格中，我们将“已婚”、“在一起”分组为“关系”
  • 而“离婚”、“寡妇”、“孤独”、“YOLO”、“荒谬”则为“单身”
In [41]:
df['Marital_Status'].value_counts()
Out[41]:
relationship    1444
Single           796
Name: Marital_Status, dtype: int64
In [42]:
# Relationship vs Single

labels_status = ["Relationship", "Single"]

count_status = pd.value_counts(df['Marital_Status'], sort=True)
count_c4.plot(kind='bar', rot=0,color=['Orange','Blue'])
plt.title("Marital Status")
plt.xticks(range(2),labels_status)
plt.xlabel("Marital Status")
plt.ylabel("Count of Marital Status")
Out[42]:
Text(0, 0.5, 'Count of Marital Status')

将不同的数据帧组合成一列以减少维数

In [44]:
df['Kids'] = df['Kidhome'] + df['Teenhome']
df['Expenses'] = df['MntWines'] + df['MntFruits'] + df['MntMeatProducts'] + df['MntFishProducts'] + df['MntSweetProducts'] + df['MntGoldProds']
df['TotalAcceptedCmp'] = df['AcceptedCmp1'] + df['AcceptedCmp2'] + df['AcceptedCmp3'] + df['AcceptedCmp4'] + df['AcceptedCmp5'] + df['Response']
df['NumTotalPurchases'] = df['NumWebPurchases'] + df['NumCatalogPurchases'] + df['NumStorePurchases'] + df['NumDealsPurchases']
In [45]:
#保存表格数据
df.to_csv('data_visuals.csv')
In [46]:
# 删除一些列以减少模型的维度和复杂性

col_del = ["AcceptedCmp1" , "AcceptedCmp2", "AcceptedCmp3" , "AcceptedCmp4","AcceptedCmp5", "Response","NumWebVisitsMonth", "NumWebPurchases","NumCatalogPurchases","NumStorePurchases","NumDealsPurchases" , "Kidhome", "Teenhome","MntWines", "MntFruits", "MntMeatProducts", "MntFishProducts", "MntSweetProducts", "MntGoldProds"]
df=df.drop(columns=col_del,axis=1)
df.head()
Out[46]:
ID Year_Birth Education Marital_Status Income Dt_Customer Recency Complain Kids Expenses TotalAcceptedCmp NumTotalPurchases
0 5524 1957 Graduation Single 58138.0 04-09-2012 58 0 0 1617 1 25
1 2174 1954 Graduation Single 46344.0 08-03-2014 38 0 2 27 0 6
2 4141 1965 Graduation relationship 71613.0 21-08-2013 26 0 0 776 0 21
3 6182 1984 Graduation relationship 26646.0 10-02-2014 26 0 1 53 0 8
4 5324 1981 PhD relationship 58293.0 19-01-2014 94 0 1 422 0 19
In [47]:
# Adding 'Age' column

df['Age'] = 2015 - df['Year_Birth']
In [48]:
df['Education'].value_counts()
Out[48]:
Graduation    1127
PhD            486
Master         370
2n Cycle       203
Basic           54
Name: Education, dtype: int64
In [49]:
# 仅将类别更改为 UG 和 PG

df['Education'] = df['Education'].replace(['PhD','2n Cycle','Graduation', 'Master'],'PG')  
df['Education'] = df['Education'].replace(['Basic'], 'UG')
In [50]:
# 客户与公司互动的天数

# 将 bt_customer 更改为时间戳格式
df['Dt_Customer'] = pd.to_datetime(df.Dt_Customer)
df['first_day'] = '01-01-2015'
df['first_day'] = pd.to_datetime(df.first_day)
df['day_engaged'] = (df['first_day'] - df['Dt_Customer']).dt.days
In [51]:
df=df.drop(columns=["ID", "Dt_Customer", "first_day", "Year_Birth", "Dt_Customer", "Recency", "Complain"],axis=1)
df.shape
Out[51]:
(2240, 9)
In [52]:
df.head()
Out[52]:
Education Marital_Status Income Kids Expenses TotalAcceptedCmp NumTotalPurchases Age day_engaged
0 PG Single 58138.0 0 1617 1 25 58 997
1 PG Single 46344.0 2 27 0 6 61 151
2 PG relationship 71613.0 0 776 0 21 50 498
3 PG relationship 26646.0 1 53 0 8 31 91
4 PG relationship 58293.0 1 422 0 19 34 347

数据可视化¶

In [53]:
fig = px.bar(df, x='Marital_Status', y='Expenses', color='Education')
fig.show()
In [54]:
fig = px.bar(df, x='Marital_Status', y='Expenses', color="Marital_Status")
fig.show()
In [55]:
# Less number of single customer
fig = px.histogram (df, x = "Expenses",  facet_row = "Marital_Status",  template = 'plotly_dark')
fig.show ()
In [56]:
fig = px.histogram (df, x = "Expenses",  facet_row = "Education",  template = 'plotly_dark')
fig.show ()
In [57]:
fig = px.histogram (df, x = "NumTotalPurchases",  facet_row = "Education",  template = 'plotly_dark')
fig.show ()
In [58]:
fig = px.histogram (df, x = "Age",  facet_row = "Marital_Status",  template = 'plotly_dark')
fig.show ()
In [59]:
fig = px.histogram (df, x = "Income",  facet_row = "Marital_Status",  template = 'plotly_dark')
fig.show ()
In [60]:
fig =  px.pie (df, names = "Marital_Status", hole = 0.4, template = "gridon")
fig.show ()
  • 35% 的客户是单身，而 64% 的客户是恋爱中的客户。
In [61]:
fig =  px.pie (df, names = "Education", hole = 0.4, template = "plotly_dark")
fig.show ()
  • 超过97%的客户来自PG背景。 和大约。 2%来自UG。
In [62]:
sns.barplot(x=df['Expenses'], y=df['Education'])
plt.title('Total expense based on the education level')
Out[62]:
Text(0.5, 1.0, 'Total expense based on the education level')
In [63]:
sns.barplot(x=df['Income'], y=df['Education'])
plt.title('Total Income based on the Education Level')
Out[63]:
Text(0.5, 1.0, 'Total Income based on the Education Level')
In [64]:
df.describe()
Out[64]:
Income Kids Expenses TotalAcceptedCmp NumTotalPurchases Age day_engaged
count 2240.000000 2240.000000 2240.000000 2240.000000 2240.000000 2240.000000 2240.000000
mean 52247.251354 0.950446 605.798214 0.446875 14.862054 46.194196 538.043304
std 25037.797168 0.751803 602.249288 0.890543 7.677173 11.984069 232.229893
min 1730.000000 0.000000 5.000000 0.000000 0.000000 19.000000 26.000000
25% 35538.750000 0.000000 68.750000 0.000000 8.000000 38.000000 366.750000
50% 51741.500000 1.000000 396.000000 0.000000 15.000000 45.000000 539.000000
75% 68289.750000 1.000000 1045.500000 1.000000 21.000000 56.000000 711.250000
max 666666.000000 3.000000 2525.000000 5.000000 44.000000 122.000000 1089.000000
In [65]:
sns.heatmap(df.corr(),annot=True)
Out[65]:
<Axes: >
In [66]:
obj = []
for i in df.columns:
    if(df[i].dtypes=="object"):
        obj.append(i)

print(obj)

['Education', 'Marital_Status']
In [67]:
# Label Encoding
from sklearn.preprocessing import LabelEncoder
In [68]:
df['Marital_Status'].value_counts()
Out[68]:
relationship    1444
Single           796
Name: Marital_Status, dtype: int64
In [69]:
lbl_encode = LabelEncoder()
for i in obj:
    df[i] = df[[i]].apply(lbl_encode.fit_transform)
In [70]:
df1 = df.copy()
df1.head()
Out[70]:
Education Marital_Status Income Kids Expenses TotalAcceptedCmp NumTotalPurchases Age day_engaged
0 0 0 58138.0 0 1617 1 25 58 997
1 0 0 46344.0 2 27 0 6 61 151
2 0 1 71613.0 0 776 0 21 50 498
3 0 1 26646.0 1 53 0 8 31 91
4 0 1 58293.0 1 422 0 19 34 347

数据标准化¶

In [71]:
from sklearn.preprocessing import StandardScaler
In [72]:
scaled_features = StandardScaler().fit_transform(df1.values)
scaled_features_df = pd.DataFrame(scaled_features, index=df1.index, columns=df1.columns)
In [73]:
scaled_features_df.head()
Out[73]:
Education Marital_Status Income Kids Expenses TotalAcceptedCmp NumTotalPurchases Age day_engaged
0 -0.157171 -1.346874 0.235327 -1.264505 1.679417 0.621248 1.320826 0.985345 1.976745
1 -0.157171 -1.346874 -0.235826 1.396361 -0.961275 -0.501912 -1.154596 1.235733 -1.667011
2 -0.157171 0.742460 0.773633 -1.264505 0.282673 -0.501912 0.799685 0.317643 -0.172468
3 -0.157171 0.742460 -1.022732 0.065928 -0.918094 -0.501912 -0.894025 -1.268149 -1.925433
4 -0.157171 0.742460 0.241519 0.065928 -0.305254 -0.501912 0.539114 -1.017761 -0.822831

肘部法则确定聚类数量¶

In [75]:
from sklearn.cluster import KMeans
In [76]:
wcss = []
for i in range(1,11):
    kmeans = KMeans(n_clusters=i, init='k-means++', random_state=42)
    kmeans.fit(scaled_features_df)
    wcss.append(kmeans.inertia_)
    # inetia_: Sum of squared distances of samples to their closest cluster center, weighted by the sample weights if provided.
plt.figure(figsize=(16,8))
plt.plot(range(1,11), wcss, 'bx-')
plt.title('The Elbow Maethod')
plt.xlabel('Number of Clusters')
plt.ylabel('WCSS')
plt.show()

拐点不清晰，肘法中并不太清楚应该选择哪个K值

剪影系数¶

In [77]:
from sklearn.metrics import silhouette_score
In [78]:
silhouette_scores = []
for i in range(2,10):
    m1 = KMeans(n_clusters=i, random_state=42)
    c = m1.fit_predict(scaled_features_df)
    silhouette_scores.append(silhouette_score(scaled_features_df, m1.fit_predict(scaled_features_df)))
    
plt.bar(range(2,10), silhouette_scores)
plt.xlabel('Number of clusters', fontsize=20)
plt.ylabel('S(i)', fontsize=20)
plt.show()
In [79]:
# 现在我们用Silhouette Score来衡量K的值
silhouette_scores
Out[79]:
[0.24145101432627075,
 0.2630066765900862,
 0.22547869857815794,
 0.2112495373878677,
 0.2149228429852001,
 0.1997135405176978,
 0.19301680336746188,
 0.19495794809915995]
In [80]:
# 获取轮廓分数的最大值并在索引中添加2 因为索引从2开始

sc = max(silhouette_scores)
num_of_clusters = silhouette_scores.index(sc)+2
print("Number of Cluster Required is: ", num_of_clusters)

Number of Cluster Required is:  3

建立模型¶

In [81]:
# 使用 K 均值算法训练预测。

kmeans = KMeans(n_clusters = num_of_clusters, random_state=42).fit(scaled_features_df)
pred = kmeans.predict(scaled_features_df)
In [82]:
pred
Out[82]:
array([1, 0, 1, ..., 1, 1, 0])
In [83]:
# 添加这些聚类值到 main dataframe (without standardization)中
df['cluster'] = pred + 1
In [84]:
df.head()
Out[84]:
Education Marital_Status Income Kids Expenses TotalAcceptedCmp NumTotalPurchases Age day_engaged cluster
0 0 0 58138.0 0 1617 1 25 58 997 2
1 0 0 46344.0 2 27 0 6 61 151 1
2 0 1 71613.0 0 776 0 21 50 498 2
3 0 1 26646.0 1 53 0 8 31 91 1
4 0 1 58293.0 1 422 0 19 34 347 1
In [85]:
# 保存数据
df.to_csv('data_visuals2.csv')
In [86]:
pl = sns.countplot(x=df["cluster"])
pl.set_title("Distribution Of The Clusters")
plt.show()

正如我们在这里看到的，与其他聚类相比，聚类 1 中的权重更大。¶

In [92]:
sns.set(rc={'axes.facecolor':'black', 'figure.facecolor':'black', 'axes.grid' : False}) 
for i in df:
    diag = sns.FacetGrid(df, col="cluster", hue="cluster", palette="Set1")
    diag.map(plt.hist, i, bins=6, ec="k") 
    diag.set_xticklabels(rotation=25, color='white')
    diag.set_yticklabels(color='white')
    diag.set_xlabels(size=16, color='white')
    diag.set_titles(size=16, color='#f01132', fontweight="bold")

报告¶

根据以上信息，我们可以将客户分为三部分：-

  1. 高度活跃客户：这些客户属于集群一。
  2. 中等活跃客户：- 这些客户属于集群二。
  3. 最不活跃的客户 ：- 这些客户属于第三集群。

高度活跃客户的特征

  • 在教育方面

    • 高活跃客户来自PG背景

  • 就婚姻状况而言

    • 恋爱关系中的人数约为。 单身人士的两倍

  • 就收入而言

    • 高度活跃客户的收入略低于中等活跃客户。

  • 就孩子而言

    • 与其他顾客相比，高度活跃的顾客拥有更多的孩子（平均 1 个孩子）。

  • 就费用而言

    • 高度活跃客户的费用低于中等客户。
    • 这些客户的平均花费。 约。 100-200单位钱。

  • 就年龄而言

    • 这些顾客的年龄在25岁至75岁之间。
    • 顾客年龄上限为 40 至 50 岁。

  • 就参与天数而言

    • 高度活跃的客户由于与公司接触的时间较长而更加忠诚。

中等活跃客户的特征

  • 在教育方面

    • 中等活跃客户也来自PG背景

  • 就婚姻状况而言

    • 与单身人士相比，恋爱中的人数略多。

  • 就收入而言

    • 中等活跃客户的收入高于其他客户。

  • 就孩子而言

    • 与高度活跃的客户相比，中等活跃的客户的子女数量较少（最多客户没有子女）。

  • 就费用而言

    • 与活跃客户相比，中等活跃客户的费用更高。
    • 这些客户的平均花费。 约。 500-2000单位钱。

  • 就年龄而言

    • 这些顾客的年龄在25岁至75岁之间。
    • 顾客年龄上限为 35 至 60 岁。

  • 就参与天数而言

    • 与高度活跃的客户相比，中等活跃的客户与公司的互动程度略低。