Novices to participate in the competition, please understand the shortcomings
Topic link: Link: https://pan.baidu.com/s/1x1U-kobvPjNMm8xnvS9Gdg
Extraction code: 7id3
Table of contents
Task 1 Data Exploration and Cleaning
Task 1.1 Data Exploration and Preprocessing
Task 2 Visual analysis of product marketing data
Task 3 Visual analysis of customer churn factors
Task 1 Data Exploration and Cleaning
Perform data exploration and cleaning on the short-term customer product purchase data "short-customer-data.csv " (referred to as short-term data) and the training set "long-customer-train.csv " (referred to as long-term data) of long-term customer resource information data.
Task 1.1 Data Exploration and Preprocessing
(1) Explore the missing values of each index data of the short-term data and the duplicate values of the " user_id " column, and delete the row data where the missing values and duplicate values are located. Please give the processing process and necessary results in the report, and the complete results are saved in the file "result1_1.xlsx ".
1. First check the data through the info function
import pandas as pd
data=pd.read_csv("F:\泰迪杯B题\B题：银行客户忠诚度分析赛题数据\B题：银行客户忠诚度分析赛题数据\short-customer-data.csv")
data.info()<class 'pandas.core.frame.DataFrame'> RangeIndex: 41176 entries, 0 to 41175 Data columns (total 14 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 user_id 41176 non-null object 1 age 41176 non-null int64 2 job 40846 non-null object 3 marital 41096 non-null object 4 education 39446 non-null object 5 default 32580 non-null object 6 housing 40186 non-null object 7 loan 40186 non-null object 8 contact 41176 non-null object 9 month 41176 non-null object 10 day_of_week 41176 non-null object 11 duration 41176 non-null int64 12 poutcome 41176 non-null object 13 y 41176 non-null object dtypes: int64(2), object(12) memory usage: 4.4+ MB
2. Delete the row data where the missing value is located
data.dropna(inplace=True)3. View duplicate values in the " user_id " column
(data["user_id"].value_counts()).head(20)BA2200001 12 BA2200775 7 BA2239485 7 BA22411696 BA2200077 2 BA2241172 2 BA2201983 2 BA2240741 2 BA2200107 2 BA2229101 1 BA2220590 1 BA2208766 1 BA2225619 1 BA2204624 1 BA2227831 1 BA2239381 1 BA2224934 1 BA2224984 1 BA2200497 1 BA2203990 1 Name: user_id, dtype: int64
4. Use the drop_duplicates function to delete the row data where the duplicate value is located
data.drop_duplicates(subset=['user_id'],inplace=True)5. Write excel file
df1=pd.DataFrame(data)
df1.to_excel('F:\\泰迪杯B题\\B题：银行客户忠诚度分析赛题数据\\result1_1.xlsx',sheet_name='sheet1',index=None)
(2) There are abnormal values of
-1 , 0 and "-" in the " Age " column
of the customer's age in the long-term data
, delete the row data in this case; the "Age " column has abnormal characters such as spaces and "years", Delete these abnormal characters but keep the age value, and store the processed value in the "Age " column. Please give the processing process and necessary results in the report, and the complete results are saved in the file "result1_2.xlsx ".
1. First take out the
"
Age
" column does not exist
data with values of
-1
,
0 and "-"
long_data=long_data[~long_data["Age"].isin([0,'-',-1])]2. Data processing
long_data["Age"] = long_data["Age"].str.replace("岁","")
long_data["Age"] = long_data["Age"].str.replace("","")long_data["Age"].unique()array(['52', '41', '42', '61', '39', '44', '53', '48', '60', '32', '38',
'51', '56', '28', '57', '45', '46', '40', '30', '59', '33', '35',
'54', '62', '23', '24', '36', '47', '73', '49', '37', '58', '21',
'55', '29', '70', '34', '43', '31', '27', '50', '66', '26', '20',
'64', '63', '71', '22', '25', '73 ', '75', '18', '1', '65', '72',
'69', '33 ', '25 ', '24 ', '76', '47 ', '74', '19', '77', '0',
'68', '49 ', '67', '63 ', '34 ', '50 ', '64 ', '29 ', '81', '27 ',
'83', '79', '78', '23 ', '30 ', '26 ', '45 ', '77 ', '84', '80',
'92', '32 ', '28 ', '31 ', '66 ', '22 ', '62 ', '46 ', '21 ',
'57 ', '82', '88'], dtype=object)
3. Write excel file
df2=pd.DataFrame(long_data)
df2.to_excel('F:\\泰迪杯B题\\B题：银行客户忠诚度分析赛题数据\\result1_2.xlsx',sheet_name='sheet1',index=None)Task 1.2 Feature Encoding
Characteristic encoding is performed on the character data in the short-term data, such as encoding the credit default situation {'no
', '
yes
'}
as
{0,1}
. Please give the processing idea, process and necessary results in the report, and the complete results will be saved in the file "result1_3.xlsx
".
1. Observe the data type
data.dtypesuser_id object age int64 job object marital object education object default object housing object loan object contact object month object day_of_week object duration int64 poutcome object y object dtype: object
2. Find the distribution of each character data value separately
data["job"].value_counts()admin. 8724 blue-collar 5670 technician 5459 services 2855 management 2310 retired 1212 self-employed 1092 entrepreneur 1088 unemployed 738 housemaid 688 student 609 Name: job, dtype: int64
data["marital"].value_counts()married 17472 single 9426 divorced 3547 Name: marital, dtype: int64
data["education"].value_counts()undergraduate 10397 postgraduate 8037 high school 7688 junior college 4312 illiterate 11 Name: education, dtype: int64
data["housing"].value_counts()yes 16498 no 13947 Name: housing, dtype: int64
data["loan"].value_counts()no 25680 yes 4765 Name: loan, dtype: int64
data["contact"].value_counts()cellular 20420 telephone 10025 Name: contact, dtype: int64
data["month"].value_counts()may 9718 jul 5075 aug 4667 jun 3608 nov 3492 apr 2112 oct 641 sep 493 mar 482 dec 157 Name: month, dtype: int64
data["day_of_week"].value_counts()thu 6388 mon 6274 wed 6116 tue 5939 fri 5728 Name: day_of_week, dtype: int64
data["poutcome"].value_counts()nonexistent 25796 failure 3458 success 1191 Name: poutcome, dtype: int64
data["y"].value_counts()no 26589 yes 3856 Name: y, dtype: int64
2. Encode the data distribution as {'no ', ' yes '} as {0,1}
data["housing"] = data["housing"].str.replace("yes","1")
data["housing"] = data["housing"].str.replace("no","0")
data["loan"] = data["loan"].str.replace("yes","1")
data["loan"] = data["loan"].str.replace("no","0")
data["default"] = data["default"].str.replace("yes","1")
data["default"] = data["default"].str.replace("no","0")
data["y"] = data["y"].str.replace("yes","1")
data["y"] = data["y"].str.replace("no","0")3. Use the one-hot encoding in sklearn to convert other non-{yes, no} specific value features into dummy variables, and assign "0" and "1" respectively
ps: There is a large amount of data, in order to avoid errors, the data is processed multiple times
from sklearn.preprocessing import OneHotEncoder
index=data.values[:,0:2]
index_data=index #获得ID列
job_marital_education=data.values[:,2:5]
enc=OneHotEncoder() #建立模型对象
df1_new=enc.fit_transform(job_marital_education).toarray() #标志转换
df1_index_education=pd.concat((pd.DataFrame(index_data),pd.DataFrame(df1_new)),axis=1) #组合为数据框
df1_index_education.columns=['user_id', 'age',
'admin','blue-collar','entrepreneur','housemaid','management','retired','self-employed','services','student',
'technician','unemployed','divorced','married','single',
'high school','illiterate','junior college','postgraduate','undergraduate']
df1_index_education.head()contact_month_week=data.values[:,8:11]
enc=OneHotEncoder() #建立模型对象
df2_new=enc.fit_transform(contact_month_week).toarray() #标志转换
df_index_week=pd.concat((pd.DataFrame(df1_index_education),pd.DataFrame(df2_new)),axis=1) #组合为数据框
df_index_week.columns=['user_id', 'age',
'admin','blue-collar','entrepreneur','housemaid','management','retired','self-employed','services','student',
'technician','unemployed','divorced','married','single',
'high school','illiterate','junior college','postgraduate','undergraduate',
'cellular','telephone','apr','aug','dec','jul','jun','mar','may','nov','oct','sep','fri','mon','thu','tue','wed']
df_index_weekdefault_loan=data.values[:,5:8]
df_now=pd.concat((pd.DataFrame(df_index_week),pd.DataFrame(default_loan)),axis=1) #组合为数据框
df_now.columns=['user_id', 'age',
'admin','blue-collar','entrepreneur','housemaid','management','retired','self-employed','services','student',
'technician','unemployed','divorced','married','single',
'high school','illiterate','junior college','postgraduate','undergraduate',
'cellular','telephone','apr','aug','dec','jul','jun','mar','may','nov','oct','sep','fri','mon','thu','tue','wed','default',
'housing','loan']
df_nowdata["y"]=data["y"].astype(dtype='int')
poutcome=data.values[:,11:]
data["y"]=data["y"].astype(dtype='int')
poutcome=data.values[:,12:]
enc=OneHotEncoder() #建立模型对象
df3_new=enc.fit_transform(poutcome).toarray() #标志转换
df_all=pd.concat((pd.DataFrame(df_index_week),pd.DataFrame(df3_new)),axis=1) #组合为数据框
df_all.columns=['user_id', 'age',
'admin','blue-collar','entrepreneur','housemaid','management','retired','self-employed','services','student',
'technician','unemployed','divorced','married','single',
'high school','illiterate','junior college','postgraduate','undergraduate',
'cellular','telephone','apr','aug','dec','jul','jun','mar','may','nov','oct','sep','fri','mon','thu','tue','wed',
'failure','nonexistent','success','1','0']
df_all.drop("0",axis=1,inplace=True)
df_all.drop("1",axis=1,inplace=True)
df_alllast=data.values[:,11:14:2]
data_last=pd.concat((pd.DataFrame(df_all),pd.DataFrame(last)),axis=1) #组合为数据框
data_last.columns=['user_id', 'age',
'admin','blue-collar','entrepreneur','housemaid','management','retired','self-employed','services','student',
'technician','unemployed','divorced','married','single',
'high school','illiterate','junior college','postgraduate','undergraduate',
'cellular','telephone','apr','aug','dec','jul','jun','mar','may','nov','oct','sep','fri','mon','thu','tue','wed','failure','nonexistent','success','duration','y']
data_lastWhen I did task 2 later, I found out that a few columns of data were missing [o(╥﹏╥)o], and then came back to make up
bb=data.values[:,5:8]
data_data=pd.concat((pd.DataFrame(data_last),pd.DataFrame(bb)),axis=1) #组合为数据框
data_data.columns=['user_id', 'age',
'admin','blue-collar','entrepreneur','housemaid','management','retired','self-employed','services','student',
'technician','unemployed','divorced','married','single',
'high school','illiterate','junior college','postgraduate','undergraduate',
'cellular','telephone','apr','aug','dec','jul','jun','mar','may','nov','oct','sep','fri','mon','thu','tue','wed',
'failure','nonexistent','success','duration','y','default','housing','loan']
data_data3. Finally write to Excel
df3=pd.DataFrame(data_data)
df3.to_excel('F:\\泰迪杯B题\\B题：银行客户忠诚度分析赛题数据\\result1_3.xlsx',sheet_name='sheet1',index=None)Mission 1 is over!
Task 2 Visual Analysis of Product Marketing Data
Based on short-term data, analyze the correlation between customers with different indicators and the behavior of purchasing bank products, and tap the loyalty of short-term customers to banks.
Task 2.1
Calculate the correlation between all indicators of short-term data, draw a heat map of the correlation coefficient, and perform necessary analysis on the results in the report.
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import missingno as ms
import plotly.express as px
import plotly.graph_objs as go
import plotly.figure_factory as ff
from plotly.subplots import make_subplots
import plotly.offline as pyo
pyo.init_notebook_mode()
sns.set_style('darkgrid')
plt.style.use('fivethirtyeight')
%matplotlib inline
import shap
plt.rc('figure',figsize=(18,9))
short_data=pd.read_excel('F:\\泰迪杯B题\\B题：银行客户忠诚度分析赛题数据\\result1_3.xlsx')corr=short_data.corr() #仅针对数值连续变量
plt.figure(figsize=(20,15))
sns.heatmap(corr
,annot=True
,linewidths=2
,linecolor='lightgrey')
plt.show()
Task 2.2
In the same canvas, draw a grouped histogram that reflects the proportion of customers of different ages under the purchase results of two products, the x-
axis is age,
the y-
axis is the proportion value, and perform necessary analysis on the results in the report.
num_20_1=0
num_30_1=0
num_40_1=0
num_50_1=0
num_60_1=0
num_70_1=0
num_80_1=0
num_90_1=0
num_100_1=0
for i in range(len(short_data["age"])):
if short_data.loc[i,'age']<=20 and short_data.loc[i,'y']==1:
num_20_1+=1
if 21<=short_data.loc[i,'age']<=30 and short_data.loc[i,'y']==1:
num_30_1+=1
if 31<=short_data.loc[i,'age']<=40 and short_data.loc[i,'y']==1:
num_40_1+=1
if 41<=short_data.loc[i,'age']<=50 and short_data.loc[i,'y']==1:
num_50_1+=1
if 51<=short_data.loc[i,'age']<=60 and short_data.loc[i,'y']==1:
num_60_1+=1
if 61<=short_data.loc[i,'age']<=70 and short_data.loc[i,'y']==1:
num_70_1+=1
if 71<=short_data.loc[i,'age']<=80 and short_data.loc[i,'y']==1:
num_80_1+=1
if 81<=short_data.loc[i,'age']<=90 and short_data.loc[i,'y']==1:
num_90_1+=1
if 91<=short_data.loc[i,'age']<=100 and short_data.loc[i,'y']==1:
num_100_1+=1
num_20_0=0
num_30_0=0
num_40_0=0
num_50_0=0
num_60_0=0
num_70_0=0
num_80_0=0
num_90_0=0
num_100_0=0
for i in range(len(short_data["age"])):
if short_data.loc[i,'age']<=20 and short_data.loc[i,'y']==0:
num_20_0+=1
if 21<=short_data.loc[i,'age']<=30 and short_data.loc[i,'y']==0:
num_30_0+=1
if 31<=short_data.loc[i,'age']<=40 and short_data.loc[i,'y']==0:
num_40_0+=1
if 41<=short_data.loc[i,'age']<=50 and short_data.loc[i,'y']==0:
num_50_0+=1
if 51<=short_data.loc[i,'age']<=60 and short_data.loc[i,'y']==0:
num_60_0+=1
if 61<=short_data.loc[i,'age']<=70 and short_data.loc[i,'y']==0:
num_70_0+=1
if 71<=short_data.loc[i,'age']<=80 and short_data.loc[i,'y']==0:
num_80_0+=1
if 81<=short_data.loc[i,'age']<=90 and short_data.loc[i,'y']==0:
num_90_0+=1
if 91<=short_data.loc[i,'age']<=100 and short_data.loc[i,'y']==0:
num_100_0+=1
total_1=num_20_1+num_30_1+num_40_1+num_50_1+num_60_1+num_70_1+num_80_1+num_90_1+num_100_1
total_0=num_20_0+num_30_0+num_40_0+num_50_0+num_60_0+num_70_0+num_80_0+num_90_0+num_100_0
from matplotlib import pyplot as plt
font = {'family': 'MicroSoft YaHei',
'weight': 'bold',
'size': '12'}
matplotlib.rc("font",**font)
x=["<20","21-30","31-40","41-50","51-60","61-70","71-80","81-90","91-100"]
y_1=[num_20_1/total_1,num_30_1/total_1,num_40_1/total_1,num_50_1/total_1,num_60_1/total_1,num_70_1/total_1,num_80_1/total_1,num_90_1/total_1,num_100_1/total_1]
y_0=[num_20_0/total_0,num_30_0/total_0,num_40_0/total_0,num_50_0/total_0,num_60_0/total_0,num_70_0/total_0,num_80_0/total_0,num_90_0/total_0,num_100_0/total_0]
ax=plt.subplot(1,2,1)
plt.bar(x,y_1,color='orange')
plt.xlabel("年龄分组")
plt.xticks(x)
plt.ylabel("人数占比")
plt.title('客户购买结果为1')
ax=plt.subplot(1,2,2)
plt.bar(x,y_0,color='green')
plt.grid(color='#95a5a6',linestyle='--',linewidth=1,axis='y',alpha=0.5)
plt.xlabel("年龄分组")
plt.xticks(x)
plt.ylabel("人数占比")
plt.title('客户购买结果为0')
plt.grid(color='#95a5a6',linestyle='--',linewidth=1,axis='y',alpha=0.5)
plt.show()
Task 2.3
In the same canvas, draw
a pie chart of product purchases of
blue-collar
and student (
student ), and set the label of the pie chart to show the proportion of product purchases.
blue_1=0
student_1=0
for i in range(len(short_data["age"])):
if short_data.loc[i,"blue-collar"]==1 and short_data.loc[i,'y']==1:
blue_1+=1
if short_data.loc[i,"student"]==1 and short_data.loc[i,'y']==1:
student_1+=1
blue_0=0
student_0=0
for i in range(len(short_data["age"])):
if short_data.loc[i,"blue-collar"]==1 and short_data.loc[i,'y']==0:
blue_0+=1
if short_data.loc[i,"student"]==1 and short_data.loc[i,'y']==0:
student_0+=1
num1=[blue_1,student_1]
num0=[blue_0,student_0]
colors = ['blue','yellow']
# #设置突出模块偏移值
expodes = (0,0)
#设置绘图属性并绘图
ax=plt.subplot(1,2,1)
plt.pie(num1,explode=expodes,labels=labels,shadow=True,colors=colors)
plt.title("产品购买结果为1")
plt.axis('equal')
ax=plt.subplot(1,2,2)
plt.pie(num0,explode=expodes,labels=labels,shadow=True,colors=colors)
plt.title("产品购买结果为0")
plt.axis('equal')
plt.show()
Task 2.4
With the product purchase results as
the x-
axis and the call duration of the visiting customers as
the y-
axis, draw a boxplot of the call duration of the visiting customers, and perform necessary analysis on the results in the report.
x=["产品购买结果为1","产品购买结果为0"]
y_1=[]
y_0=[]
for i in range(len(short_data["age"])):
if short_data.loc[i,'y']==0:
y_1.append(short_data.loc[i,'duration'])
if short_data.loc[i,'y']==1:
y_0.append(short_data.loc[i,'duration'])
dt = pd.DataFrame({'产品购买结果为0': y_0})
dt = pd.DataFrame({'产品购买结果为1': y_1})
dt.boxplot() #对数据框中每列画箱线图，pandas自己有处理的过程
plt.show()
Task 3 Visual analysis of customer churn factors
Based on long-term data analysis, the factors that lead to bank customer loss are presented visually.
Task 3.1
In the same canvas, draw a line chart reflecting the proportion of customers of different ages under two churn situations, the x-
axis is the age, and
the y-
axis is the proportion value
import pandas as pd
long_data=pd.read_excel('F:\\泰迪杯B题\\B题：银行客户忠诚度分析赛题数据\\result1_2.xlsx')
num_20_1=0
num_30_1=0
num_40_1=0
num_50_1=0
num_60_1=0
num_70_1=0
num_80_1=0
num_90_1=0
num_100_1=0
for i in range(len(short_data["Age"])):
if short_data.loc[i,'Age']<=20 and short_data.loc[i,'Exited']==1:
num_20_1+=1
if 21<=short_data.loc[i,'Age']<=30 and short_data.loc[i,'Exited']==1:
num_30_1+=1
if 31<=short_data.loc[i,'Age']<=40 and short_data.loc[i,'Exited']==1:
num_40_1+=1
if 41<=short_data.loc[i,'Age']<=50 and short_data.loc[i,'Exited']==1:
num_50_1+=1
if 51<=short_data.loc[i,'Age']<=60 and short_data.loc[i,'Exited']==1:
num_60_1+=1
if 61<=short_data.loc[i,'Age']<=70 and short_data.loc[i,'Exited']==1:
num_70_1+=1
if 71<=short_data.loc[i,'Age']<=80 and short_data.loc[i,'Exited']==1:
num_80_1+=1
if 81<=short_data.loc[i,'Age']<=90 and short_data.loc[i,'Exited']==1:
num_90_1+=1
if 91<=short_data.loc[i,'Age']<=100 and short_data.loc[i,'Exited']==1:
num_100_1+=1
num_20_0=0
num_30_0=0
num_40_0=0
num_50_0=0
num_60_0=0
num_70_0=0
num_80_0=0
num_90_0=0
num_100_0=0
for i in range(len(short_data["Age"])):
if short_data.loc[i,'Age']<=20 and short_data.loc[i,'Exited']==0:
num_20_0+=1
if 21<=short_data.loc[i,'Age']<=30 and short_data.loc[i,'Exited']==0:
num_30_0+=1
if 31<=short_data.loc[i,'Age']<=40 and short_data.loc[i,'Exited']==0:
num_40_0+=1
if 41<=short_data.loc[i,'Age']<=50 and short_data.loc[i,'Exited']==0:
num_50_0+=1
if 51<=short_data.loc[i,'Age']<=60 and short_data.loc[i,'Exited']==0:
num_60_0+=1
if 61<=short_data.loc[i,'Age']<=70 and short_data.loc[i,'Exited']==0:
num_70_0+=1
if 71<=short_data.loc[i,'Age']<=80 and short_data.loc[i,'Exited']==0:
num_80_0+=1
if 81<=short_data.loc[i,'Age']<=90 and short_data.loc[i,'Exited']==0:
num_90_0+=1
if 91<=short_data.loc[i,'Age']<=100 and short_data.loc[i,'Exited']==0:
num_100_0+=1
total_1=num_20_1+num_30_1+num_40_1+num_50_1+num_60_1+num_70_1+num_80_1+num_90_1+num_100_1
total_1
total_0=num_20_0+num_30_0+num_40_0+num_50_0+num_60_0+num_70_0+num_80_0+num_90_0+num_100_0
total_0
import pandas as pd
import matplotlib
from matplotlib import pyplot as plt
font = {'family': 'MicroSoft YaHei',
'weight': 'bold',
'size': '12'}
matplotlib.rc("font",**font)
x=["<20","21-30","31-40","41-50","51-60","61-70","71-80","81-90","91-100"]
y_1=[num_20_1/total_1,num_30_1/total_1,num_40_1/total_1,num_50_1/total_1,num_60_1/total_1,num_70_1/total_1,num_80_1/total_1,num_90_1/total_1,num_100_1/total_1]
y_0=[num_20_0/total_0,num_30_0/total_0,num_40_0/total_0,num_50_0/total_0,num_60_0/total_0,num_70_0/total_0,num_80_0/total_0,num_90_0/total_0,num_100_0/total_0]
plt.figure(figsize=(12,5))
ax=plt.subplot(1,2,1)
plt.plot(x,y_1,color='orange')
plt.xlabel("年龄分组")
plt.xticks(x)
plt.ylabel("人数占比")
plt.title('客户流失')
ax=plt.subplot(1,2,2)
plt.plot(x,y_0,color='green')
plt.grid(color='#95a5a6',linestyle='--',linewidth=1,axis='y',alpha=0.5)
plt.xlabel("年龄分组")
plt.xticks(x)
plt.ylabel("人数占比")
plt.title('客户不流失')
plt.grid(color='#95a5a6',linestyle='--',linewidth=1,axis='y',alpha=0.5)
plt.show()
Task 3.2
In the same canvas, draw a scatter plot reflecting the customer credit qualification and age distribution under two churn scenarios, with
age on the
x- axis and credit qualification on
the y- axis.
x_1=[]
y_1=[]
x_0=[]
y_0=[]
for i in range(len(short_data["Age"])):
if short_data.loc[i,'Exited']==1:
x1=short_data.loc[i,'Age']
y1=short_data.loc[i,'CreditScore']
x_1.append(x1)
y_1.append(y1)
if short_data.loc[i,'Exited']==0:
x0=short_data.loc[i,'Age']
y0=short_data.loc[i,'CreditScore']
x_0.append(x0)
y_0.append(y0)
fig=plt.figure(figsize=(12,5),dpi=100)
ax=plt.subplot(1,2,1)
plt.scatter(x_1,y_1,s=1,)
plt.xlabel("年龄")
plt.ylabel("信用资格")
plt.title('客户流失')
ax=plt.subplot(1,2,2)
plt.scatter(x_0,y_0,s=1,)
plt.xlabel("年龄")
# plt.ylabel("信用资格")
plt.yticks([])
plt.title('客户不流失')
plt.show()
Task 3.3
Construct a perspective table containing the proportion of customers of each account age under different churn conditions (see Table 4 for details
), and draw a stacked column chart reflecting the proportion of account age of customers in the two churn situations on the same canvas, The x-
axis is the age of the customer, and
the y-
axis is the proportion.
count_0_0=0
count_1_0=0
count_2_0=0
count_3_0=0
count_4_0=0
count_5_0=0
count_6_0=0
count_7_0=0
count_8_0=0
count_9_0=0
count_10_0=0
count_0=[count_0_0,count_1_0,count_2_0,count_3_0,count_4_0,count_5_0,count_6_0,count_7_0,count_8_0,count_9_0,count_10_0]
count_0_1=0
count_1_1=0
count_2_1=0
count_3_1=0
count_4_1=0
count_5_1=0
count_6_1=0
count_7_1=0
count_8_1=0
count_9_1=0
count_10_1=0
count_1=[count_0_1,count_1_1,count_2_1,count_3_1,count_4_1,count_5_1,count_6_1,count_7_1,count_8_1,count_9_1,count_10_1]
for i in range(len(short_data["Age"])):
if short_data.loc[i,'Exited']==1:
a1=short_data.loc[i,'Tenure']
if a1==0:
count_0_1+=1
if a1==1:
count_1_1+=1
if a1==2:
count_2_1+=1
if a1==3:
count_3_1+=1
if a1==4:
count_4_1+=1
if a1==5:
count_5_1+=1
if a1==6:
count_6_1+=1
if a1==7:
count_7_1+=1
if a1==8:
count_8_1+=1
if a1==9:
count_9_1+=1
if a1==10:
count_10_1+=1
if short_data.loc[i,'Exited']==0:
a2=short_data.loc[i,'Tenure']
if a2==0:
count_0_0+=1
if a2==1:
count_1_0+=1
if a2==2:
count_2_0+=1
if a2==3:
count_3_0+=1
if a2==4:
count_4_0+=1
if a2==5:
count_5_0+=1
if a2==6:
count_6_0+=1
if a2==7:
count_7_0+=1
if a2==8:
count_8_0+=1
if a2==9:
count_9_0+=1
if a2==10:
count_10_0+=1Get the pivot table:
total_0=0
total_1=0
for i in range(0,11):
total_0+=count_0[i]
total_1+=count_1[i]
total_0
total_1
for i in range(0,11):
count_0[i]=count_0[i]/total_0
count_1[i]=count_1[i]/total_1
data_1={'count_0':count_0,'count_1':count_1}
df_1=pd.DataFrame(data_1)
df_1.plot(kind='bar',stacked=True,alpha=0.5) 
Task 3.4
According to Table
5
and Table
6
, account age and customer financial assets are divided, and feature coding is performed as new customer features, in which the customer status is stored in the "Status" column, the asset
stage is stored in the "
AssetStage
" column, and the coding results are saved into the file "result3.xlsx
".
Status=[]
AssetStage=[]
for i in range(len(short_data["Age"])):
if 0<=short_data.loc[i,'Tenure']<=3:
Status[i]="新客户"
if 3<short_data.loc[i,'Tenure']<=6:
Status[i]="稳定客户"
if short_data.loc[i,'Tenure']>6:
Status[i]="老客户"
for i in range(len(short_data["Age"])):
if 0<=short_data.loc[i,'Balance']<=50000:
AssetStage[i]="低资产"
if 50000<short_data.loc[i,'Balance']<=90000:
AssetStage[i]="中下资产"
if 90000<short_data.loc[i,'Balance']<=120000:
AssetStage[i]="中上资产"
if short_data.loc[i,'Balance']>120000:
AssetStage[i]="高资产"
data_1={'Status':Status,'AssetStage':AssetStage}
df_1=pd.DataFrame(data_1)
df_1.to_excel('F:\\泰迪杯B题\\B题：银行客户忠诚度分析赛题数据\\result3.xlsx',sheet_name='sheet1',index=None)
(2)
Count the number of customers lost by new and old customers in each asset stage, draw a heat map on the same canvas, set the maximum and minimum values of the color of the heat map to 1300 and 100, and perform necessary analysis on the results
in
the
report .
test=pd.read_excel('F:\\泰迪杯B题\\B题：银行客户忠诚度分析赛题数据\\任务3.4.2.xlsx')
new_lower=0
new_low=0
new_high=0
new_higher=0
old_lower=0
old_low=0
old_high=0
old_higher=0
for i in range(len(test['Exited'])):
if test.loc[i,'Exited']==1 and test.loc[i,'Status']=='新客户'and test.loc[i,'AssetStage']=="低资产":
new_lower+=1
if test.loc[i,'Exited']==1 and test.loc[i,'Status']=='新客户'and test.loc[i,'AssetStage']=="中下资产":
new_low+=1
if test.loc[i,'Exited']==1 and test.loc[i,'Status']=='新客户'and test.loc[i,'AssetStage']=="中上资产":
new_high+=1
if test.loc[i,'Exited']==1 and test.loc[i,'Status']=='新客户'and test.loc[i,'AssetStage']=="高资产":
new_higher+=1
if test.loc[i,'Exited']==0 and test.loc[i,'Status']=='老客户'and test.loc[i,'AssetStage']=="低资产":
old_lower+=1
if test.loc[i,'Exited']==0 and test.loc[i,'Status']=='老客户'and test.loc[i,'AssetStage']=="中下资产":
old_low+=1
if test.loc[i,'Exited']==0 and test.loc[i,'Status']=='老客户'and test.loc[i,'AssetStage']=="中上资产":
old_high+=1
if test.loc[i,'Exited']==0 and test.loc[i,'Status']=='老客户'and test.loc[i,'AssetStage']=="高资产":
old_higher+=1
new=[new_lower,new_low,new_high,new_higher]
old=[old_lower,old_low,old_high,old_higher]
data_2={'new_lower':[180],'new_low':[57],'new_high':[181],'new_higher':[246],
'old_lower':[1066],'old_low':[234],'old_high':[536],'old_higher':[800]}
df_2=pd.DataFrame(data_2)
plt.figure(figsize=(10,5))
sns.heatmap(df_2,vmax=1300,vmin=100)
ps: Since the space is too long, let's write it here for the time being. The rest will be written in another article, and then I will make a summary of the competition
2022 The 5th "Teddy Cup" Data Analysis Skills Competition - Question B - Bank Customer Loyalty Analysis (Part 2) Link: https://blog.csdn.net/weixin_60200880/article/details/127939604?spm=1001.2014.3001.5502