4/9/23, 6:41 PM EDA - Jupyter Notebook

Prepared by Asif Bhat

Exploratory Data Analysis ¶

In [1]: import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import plotly.express as px
import plotly as ply
import seaborn as sns
import warnings
import plotly.graph_objects as go
import plotly.offline as po
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, ipl
import plotly.express as px
import plotly.figure_factory as ff
from plotly.subplots import make_subplots
warnings.filterwarnings('ignore')
import plotly.io as pio
pio.renderers.default = 'iframe'
pio.templates.default = "plotly_dark"

Explore Dataset

In [2]: app_data = pd.read_csv('application_data.csv')

app_data.head()

Out[2]:
SK_ID_CURR TARGET NAME_CONTRACT_TYPE CODE_GENDER FLAG_OWN_CAR FLAG_OWN_R

0 100002 1 Cash loans M N

1 100003 0 Cash loans F N

2 100004 0 Revolving loans M Y

3 100006 0 Cash loans F N

4 100007 0 Cash loans M N

5 rows × 122 columns

In [3]: app_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 307511 entries, 0 to 307510
Columns: 122 entries, SK_ID_CURR to AMT_REQ_CREDIT_BUREAU_YEAR
dtypes: float64(65), int64(41), object(16)
memory usage: 286.2+ MB

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 1/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [4]: app_data.shape # 122 Columns

Out[4]: (307511, 122)

In [5]: # Summary of numeric columns

app_data.describe()

Out[5]:
SK_ID_CURR TARGET CNT_CHILDREN AMT_INCOME_TOTAL AMT_CREDIT AMT_AN

count 307511.000000 307511.000000 307511.000000 3.075110e+05 3.075110e+05 307499.0

mean 278180.518577 0.080729 0.417052 1.687979e+05 5.990260e+05 27108.5

std 102790.175348 0.272419 0.722121 2.371231e+05 4.024908e+05 14493.7

min 100002.000000 0.000000 0.000000 2.565000e+04 4.500000e+04 1615.5

25% 189145.500000 0.000000 0.000000 1.125000e+05 2.700000e+05 16524.0

50% 278202.000000 0.000000 0.000000 1.471500e+05 5.135310e+05 24903.0

75% 367142.500000 0.000000 1.000000 2.025000e+05 8.086500e+05 34596.0

max 456255.000000 1.000000 19.000000 1.170000e+08 4.050000e+06 258025.5

8 rows × 106 columns

In [6]: # Most of the columns are of type integer or float.

app_data.dtypes.value_counts()

Out[6]: float64 65
int64 41
object 16
dtype: int64

Data Cleaning

Dropping Columns with high percentage of NULL values

In [7]: # Percentage of NULL Values in descending order

(app_data.isnull().mean()*100).sort_values(ascending=False)

Out[7]: COMMONAREA_MEDI 69.872297

COMMONAREA_AVG 69.872297
COMMONAREA_MODE 69.872297
NONLIVINGAPARTMENTS_MODE 69.432963
NONLIVINGAPARTMENTS_AVG 69.432963
...
NAME_HOUSING_TYPE 0.000000
NAME_FAMILY_STATUS 0.000000
NAME_EDUCATION_TYPE 0.000000
NAME_INCOME_TYPE 0.000000
SK_ID_CURR 0.000000
Length: 122, dtype: float64

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 2/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [8]: # Columns with NULL Values greater than 40%

s1= (app_data.isnull().mean()*100).sort_values(ascending=False)[app_data.is
s1

Out[8]: COMMONAREA_MEDI 69.872297

COMMONAREA_AVG 69.872297
COMMONAREA_MODE 69.872297
NONLIVINGAPARTMENTS_MODE 69.432963
NONLIVINGAPARTMENTS_AVG 69.432963
NONLIVINGAPARTMENTS_MEDI 69.432963
FONDKAPREMONT_MODE 68.386172
LIVINGAPARTMENTS_MODE 68.354953
LIVINGAPARTMENTS_AVG 68.354953
LIVINGAPARTMENTS_MEDI 68.354953
FLOORSMIN_AVG 67.848630
FLOORSMIN_MODE 67.848630
FLOORSMIN_MEDI 67.848630
YEARS_BUILD_MEDI 66.497784
YEARS_BUILD_MODE 66.497784
YEARS_BUILD_AVG 66.497784
OWN_CAR_AGE 65.990810
LANDAREA_MEDI 59.376738
LANDAREA_MODE 59.376738
LANDAREA_AVG 59.376738
BASEMENTAREA_MEDI 58.515956
BASEMENTAREA_AVG 58.515956
BASEMENTAREA_MODE 58.515956
EXT_SOURCE_1 56.381073
NONLIVINGAREA_MODE 55.179164
NONLIVINGAREA_AVG 55.179164
NONLIVINGAREA_MEDI 55.179164
ELEVATORS_MEDI 53.295980
ELEVATORS_AVG 53.295980
ELEVATORS_MODE 53.295980
WALLSMATERIAL_MODE 50.840783
APARTMENTS_MEDI 50.749729
APARTMENTS_AVG 50.749729
APARTMENTS_MODE 50.749729
ENTRANCES_MEDI 50.348768
ENTRANCES_AVG 50.348768
ENTRANCES_MODE 50.348768
LIVINGAREA_AVG 50.193326
LIVINGAREA_MODE 50.193326
LIVINGAREA_MEDI 50.193326
HOUSETYPE_MODE 50.176091
FLOORSMAX_MODE 49.760822
FLOORSMAX_MEDI 49.760822
FLOORSMAX_AVG 49.760822
YEARS_BEGINEXPLUATATION_MODE 48.781019
YEARS_BEGINEXPLUATATION_MEDI 48.781019
YEARS_BEGINEXPLUATATION_AVG 48.781019
TOTALAREA_MODE 48.268517
EMERGENCYSTATE_MODE 47.398304
dtype: float64

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 3/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [9]: fig= px.bar(data_frame=s1,

x=s1.index.tolist(),
y=s1.values,
color=s1.values,
text=s1.values.round()
)
fig.update_traces(textposition='outside',marker_coloraxis=None)
fig.update_xaxes(title='Columns')
fig.update_yaxes(title='Percentage')
fig.update_layout(
title=dict(text = "Null Value Percentage",x=0.5,y=0.95)
title_font_size=20,
showlegend=False,
height =600,
)
fig.show()

Null Value Percentage

70 70 70 69 69 69
70 68 68 68 68 68 68 68
66 66 66 66

59 59 59 59 59 59
60 56 55 55 55
53 53 53
51 51
50
Percentage

40

30

20

10

0
COMMONAREA_MEDI
COMMONAREA_AVG
COMMONAREA_MODE
NONLIVINGAPARTMENTS_MODE
NONLIVINGAPARTMENTS_AVG
NONLIVINGAPARTMENTS_MEDI
FONDKAPREMONT_MODE
LIVINGAPARTMENTS_MODE
LIVINGAPARTMENTS_AVG
LIVINGAPARTMENTS_MEDI
FLOORSMIN_AVG
FLOORSMIN_MODE
FLOORSMIN_MEDI
YEARS_BUILD_MEDI
YEARS_BUILD_MODE
YEARS_BUILD_AVG
OWN_CAR_AGE
LANDAREA_MEDI
LANDAREA_MODE
LANDAREA_AVG
BASEMENTAREA_MEDI
BASEMENTAREA_AVG
BASEMENTAREA_MODE
EXT_SOURCE_1
NONLIVINGAREA_MODE
NONLIVINGAREA_AVG
NONLIVINGAREA_MEDI
ELEVATORS_MEDI
ELEVATORS_AVG
ELEVATORS_MODE
WALLSMATERIAL_MODE
APARTMENTS_MEDI

Columns

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 4/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [10]: # Get Column names with NULL percentage greater than 40%
cols = (app_data.isnull().mean()*100 > 40)[app_data.isnull().mean()*100 > 4
cols

Out[10]: ['OWN_CAR_AGE',
'EXT_SOURCE_1',
'APARTMENTS_AVG',
'BASEMENTAREA_AVG',
'YEARS_BEGINEXPLUATATION_AVG',
'YEARS_BUILD_AVG',
'COMMONAREA_AVG',
'ELEVATORS_AVG',
'ENTRANCES_AVG',
'FLOORSMAX_AVG',
'FLOORSMIN_AVG',
'LANDAREA_AVG',
'LIVINGAPARTMENTS_AVG',
'LIVINGAREA_AVG',
'NONLIVINGAPARTMENTS_AVG',
'NONLIVINGAREA_AVG',
'APARTMENTS_MODE',
'BASEMENTAREA_MODE',
'YEARS_BEGINEXPLUATATION_MODE',
'YEARS_BUILD_MODE',
'COMMONAREA_MODE',
'ELEVATORS_MODE',
'ENTRANCES_MODE',
'FLOORSMAX_MODE',
'FLOORSMIN_MODE',
'LANDAREA_MODE',
'LIVINGAPARTMENTS_MODE',
'LIVINGAREA_MODE',
'NONLIVINGAPARTMENTS_MODE',
'NONLIVINGAREA_MODE',
'APARTMENTS_MEDI',
'BASEMENTAREA_MEDI',
'YEARS_BEGINEXPLUATATION_MEDI',
'YEARS_BUILD_MEDI',
'COMMONAREA_MEDI',
'ELEVATORS_MEDI',
'ENTRANCES_MEDI',
'FLOORSMAX_MEDI',
'FLOORSMIN_MEDI',
'LANDAREA_MEDI',
'LIVINGAPARTMENTS_MEDI',
'LIVINGAREA_MEDI',
'NONLIVINGAPARTMENTS_MEDI',
'NONLIVINGAREA_MEDI',
'FONDKAPREMONT_MODE',
'HOUSETYPE_MODE',
'TOTALAREA_MODE',
'WALLSMATERIAL_MODE',
'EMERGENCYSTATE_MODE']

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 5/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [11]: # We are good to delete 49 columns because NULL percentage for these column
len(cols)

Out[11]: 49

In [12]: # Drop 49 columns

app_data.drop(columns=cols,inplace=True)

In [13]: app_data.shape # 307511 rows & 73 Columns

Out[13]: (307511, 73)

In [14]: # NULL Values percentage in new dataset

s2= (app_data.isnull().mean()*100).sort_values(ascending=False)
s2

Out[14]: OCCUPATION_TYPE 31.345545

EXT_SOURCE_3 19.825307
AMT_REQ_CREDIT_BUREAU_YEAR 13.501631
AMT_REQ_CREDIT_BUREAU_QRT 13.501631
AMT_REQ_CREDIT_BUREAU_MON 13.501631
...
REG_REGION_NOT_LIVE_REGION 0.000000
REG_REGION_NOT_WORK_REGION 0.000000
LIVE_REGION_NOT_WORK_REGION 0.000000
TARGET 0.000000
REG_CITY_NOT_LIVE_CITY 0.000000
Length: 73, dtype: float64

In [15]: s2.head(10)

Out[15]: OCCUPATION_TYPE 31.345545

EXT_SOURCE_3 19.825307
AMT_REQ_CREDIT_BUREAU_YEAR 13.501631
AMT_REQ_CREDIT_BUREAU_QRT 13.501631
AMT_REQ_CREDIT_BUREAU_MON 13.501631
AMT_REQ_CREDIT_BUREAU_WEEK 13.501631
AMT_REQ_CREDIT_BUREAU_DAY 13.501631
AMT_REQ_CREDIT_BUREAU_HOUR 13.501631
NAME_TYPE_SUITE 0.420148
OBS_30_CNT_SOCIAL_CIRCLE 0.332021
dtype: float64

Imputation of Missing Values

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 6/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [16]: app_data.head()

Out[16]:
SK_ID_CURR TARGET NAME_CONTRACT_TYPE CODE_GENDER FLAG_OWN_CAR FLAG_OWN_R

0 100002 1 Cash loans M N

1 100003 0 Cash loans F N

2 100004 0 Revolving loans M Y

3 100006 0 Cash loans F N

4 100007 0 Cash loans M N

5 rows × 73 columns

In [17]: '''
Impute the missing values of below columns with mode
- AMT_REQ_CREDIT_BUREAU_MONTH
- AMT_REQ_CREDIT_BUREAU_WEEK
- AMT_REQ_CREDIT_BUREAU_DAY
- AMT_REQ_CREDIT_BUREAU_HOUR
- AMT_REQ_CREDIT_BUREAU_QRT
'''
for i in s2.head(10).index.to_list():
if 'AMT_REQ_CREDIT' in i:
print('Most frequent value in {0} is : {1}'.format(i,app_data[i].mo
print('Imputing the missing value with : {0}'.format(app_data[i].mo
app_data[i].fillna(app_data[i].mode()[0],inplace=True)
print('NULL Values in {0} after imputation : {1}'.format(i,app_data
print()

Most frequent value in AMT_REQ_CREDIT_BUREAU_YEAR is : 0.0

Imputing the missing value with : 0.0
NULL Values in AMT_REQ_CREDIT_BUREAU_YEAR after imputation : 0

Most frequent value in AMT_REQ_CREDIT_BUREAU_QRT is : 0.0

Imputing the missing value with : 0.0
NULL Values in AMT_REQ_CREDIT_BUREAU_QRT after imputation : 0

Most frequent value in AMT_REQ_CREDIT_BUREAU_MON is : 0.0

Imputing the missing value with : 0.0
NULL Values in AMT_REQ_CREDIT_BUREAU_MON after imputation : 0

Most frequent value in AMT_REQ_CREDIT_BUREAU_WEEK is : 0.0

Imputing the missing value with : 0.0
NULL Values in AMT_REQ_CREDIT_BUREAU_WEEK after imputation : 0

Most frequent value in AMT_REQ_CREDIT_BUREAU_DAY is : 0.0

Imputing the missing value with : 0.0
NULL Values in AMT_REQ_CREDIT_BUREAU_DAY after imputation : 0

Most frequent value in AMT_REQ_CREDIT_BUREAU_HOUR is : 0.0

Imputing the missing value with : 0.0
NULL Values in AMT_REQ_CREDIT_BUREAU_HOUR after imputation : 0

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 7/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [18]: # Missing value percentage of remaining columns

(app_data.isnull().mean()*100).sort_values(ascending=False)

Out[18]: OCCUPATION_TYPE 31.345545

EXT_SOURCE_3 19.825307
NAME_TYPE_SUITE 0.420148
OBS_30_CNT_SOCIAL_CIRCLE 0.332021
DEF_30_CNT_SOCIAL_CIRCLE 0.332021
...
REG_REGION_NOT_LIVE_REGION 0.000000
REG_REGION_NOT_WORK_REGION 0.000000
LIVE_REGION_NOT_WORK_REGION 0.000000
TARGET 0.000000
AMT_REQ_CREDIT_BUREAU_YEAR 0.000000
Length: 73, dtype: float64

Impute missing values for OCCUPATION_TYPE

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 8/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [19]: # We can impute missing values in 'OCCUPATION_TYPE' column with 'Laborers'

fig=px.bar(app_data.OCCUPATION_TYPE.value_counts(),color=app_data.OCCUPATIO
fig.update_traces(textposition='outside',marker_coloraxis=None)
fig.update_xaxes(title='Occupation Type')
fig.update_yaxes(title='Count')
fig.update_layout(
title=dict(text = "Occupation Type Frequency",x=0.5,y=0
title_font_size=20,
showlegend=False,
height =450,
)
fig.show()

Occupation Type Frequenc

50k

40k
Count

30k

20k

10k

0
La Sa Co Ma Dr Hi Ac Me Se Co Cle Pr
bo les re na ive gh co dic cu ok an iv
re sta sta ger rs sk un i ne r i ty i n i ng
rs ff s ill tan gs
ff tec ts sta sta taf sta
hs ff ff f
taf
f

Occupation Type

In [20]: app_data.OCCUPATION_TYPE.fillna('Laborers',inplace=True)

Impute Missing values (XNA) in CODE_GENDER with mode

In [21]: app_data['CODE_GENDER'].value_counts()

Out[21]: F 202448
M 105059
XNA 4
Name: CODE_GENDER, dtype: int64

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 9/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [22]: app_data['CODE_GENDER'].replace(to_replace='XNA',value=app_data['CODE_GENDE

In [23]: app_data['CODE_GENDER'].value_counts()

Out[23]: F 202452
M 105059
Name: CODE_GENDER, dtype: int64

Impute missing values for EXT_SOURCE_3

In [24]: app_data.EXT_SOURCE_3.dtype

Out[24]: dtype('float64')

In [25]: app_data.EXT_SOURCE_3.fillna(app_data.EXT_SOURCE_3.median(),inplace=True)

In [26]: # Percentage of missing values after imputation

(app_data.isnull().mean()*100).sort_values(ascending=False)

Out[26]: NAME_TYPE_SUITE 0.420148

DEF_60_CNT_SOCIAL_CIRCLE 0.332021
OBS_30_CNT_SOCIAL_CIRCLE 0.332021
DEF_30_CNT_SOCIAL_CIRCLE 0.332021
OBS_60_CNT_SOCIAL_CIRCLE 0.332021
...
REG_REGION_NOT_LIVE_REGION 0.000000
REG_REGION_NOT_WORK_REGION 0.000000
LIVE_REGION_NOT_WORK_REGION 0.000000
TARGET 0.000000
AMT_REQ_CREDIT_BUREAU_YEAR 0.000000
Length: 73, dtype: float64

In [27]: # Replace 'XNA' with NaN

app_data = app_data.replace('XNA',np.NaN)

DELETE all flag columns

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 10/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [28]: app_data.columns

Out[28]: Index(['SK_ID_CURR', 'TARGET', 'NAME_CONTRACT_TYPE', 'CODE_GENDER',

'FLAG_OWN_CAR', 'FLAG_OWN_REALTY', 'CNT_CHILDREN', 'AMT_INCOME_TOT
AL',
'AMT_CREDIT', 'AMT_ANNUITY', 'AMT_GOODS_PRICE', 'NAME_TYPE_SUITE',
'NAME_INCOME_TYPE', 'NAME_EDUCATION_TYPE', 'NAME_FAMILY_STATUS',
'NAME_HOUSING_TYPE', 'REGION_POPULATION_RELATIVE', 'DAYS_BIRTH',
'DAYS_EMPLOYED', 'DAYS_REGISTRATION', 'DAYS_ID_PUBLISH', 'FLAG_MOB
IL',
'FLAG_EMP_PHONE', 'FLAG_WORK_PHONE', 'FLAG_CONT_MOBILE', 'FLAG_PHO
NE',
'FLAG_EMAIL', 'OCCUPATION_TYPE', 'CNT_FAM_MEMBERS',
'REGION_RATING_CLIENT', 'REGION_RATING_CLIENT_W_CITY',
'WEEKDAY_APPR_PROCESS_START', 'HOUR_APPR_PROCESS_START',
'REG_REGION_NOT_LIVE_REGION', 'REG_REGION_NOT_WORK_REGION',
'LIVE_REGION_NOT_WORK_REGION', 'REG_CITY_NOT_LIVE_CITY',
'REG_CITY_NOT_WORK_CITY', 'LIVE_CITY_NOT_WORK_CITY',
'ORGANIZATION_TYPE', 'EXT_SOURCE_2', 'EXT_SOURCE_3',
'OBS_30_CNT_SOCIAL_CIRCLE', 'DEF_30_CNT_SOCIAL_CIRCLE',
'OBS_60_CNT_SOCIAL_CIRCLE', 'DEF_60_CNT_SOCIAL_CIRCLE',
'DAYS_LAST_PHONE_CHANGE', 'FLAG_DOCUMENT_2', 'FLAG_DOCUMENT_3',
'FLAG_DOCUMENT_4', 'FLAG_DOCUMENT_5', 'FLAG_DOCUMENT_6',
'FLAG_DOCUMENT_7', 'FLAG_DOCUMENT_8', 'FLAG_DOCUMENT_9',
'FLAG_DOCUMENT_10', 'FLAG_DOCUMENT_11', 'FLAG_DOCUMENT_12',
'FLAG_DOCUMENT_13', 'FLAG_DOCUMENT_14', 'FLAG_DOCUMENT_15',
'FLAG_DOCUMENT_16', 'FLAG_DOCUMENT_17', 'FLAG_DOCUMENT_18',
'FLAG_DOCUMENT_19', 'FLAG_DOCUMENT_20', 'FLAG_DOCUMENT_21',
'AMT_REQ_CREDIT_BUREAU_HOUR', 'AMT_REQ_CREDIT_BUREAU_DAY',
'AMT_REQ_CREDIT_BUREAU_WEEK', 'AMT_REQ_CREDIT_BUREAU_MON',
'AMT_REQ_CREDIT_BUREAU_QRT', 'AMT_REQ_CREDIT_BUREAU_YEAR'],
dtype='object')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 11/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [29]: # Flag Columns

col =[]
for i in app_data.columns:
if 'FLAG' in i:
col.append(i)
col

Out[29]: ['FLAG_OWN_CAR',
'FLAG_OWN_REALTY',
'FLAG_MOBIL',
'FLAG_EMP_PHONE',
'FLAG_WORK_PHONE',
'FLAG_CONT_MOBILE',
'FLAG_PHONE',
'FLAG_EMAIL',
'FLAG_DOCUMENT_2',
'FLAG_DOCUMENT_3',
'FLAG_DOCUMENT_4',
'FLAG_DOCUMENT_5',
'FLAG_DOCUMENT_6',
'FLAG_DOCUMENT_7',
'FLAG_DOCUMENT_8',
'FLAG_DOCUMENT_9',
'FLAG_DOCUMENT_10',
'FLAG_DOCUMENT_11',
'FLAG_DOCUMENT_12',
'FLAG_DOCUMENT_13',
'FLAG_DOCUMENT_14',
'FLAG_DOCUMENT_15',
'FLAG_DOCUMENT_16',
'FLAG_DOCUMENT_17',
'FLAG_DOCUMENT_18',
'FLAG_DOCUMENT_19',
'FLAG_DOCUMENT_20',
'FLAG_DOCUMENT_21']

In [30]: # DELETE all flag columns as they won't be much useful in our analysis
app_data.drop(columns=col,inplace=True)
app_data.head()
​
#OR
​
#app_data= app_data[[i for i in app_data.columns if 'FLAG' not in i]]

Out[30]:
SK_ID_CURR TARGET NAME_CONTRACT_TYPE CODE_GENDER CNT_CHILDREN AMT_INCOME_

0 100002 1 Cash loans M 0 2

1 100003 0 Cash loans F 0 2

2 100004 0 Revolving loans M 0

3 100006 0 Cash loans F 0 1

4 100007 0 Cash loans M 0 1

5 rows × 45 columns

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 12/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Impute Missing values for AMT_ANNUITY & AMT_GOODS_PRICE

In [31]: col=['AMT_INCOME_TOTAL','AMT_CREDIT', 'AMT_ANNUITY', 'AMT_GOODS_PRICE']

for i in col:
print('Null Values in {0} : {1}'.format(i,app_data[i].isnull().sum()))

Null Values in AMT_INCOME_TOTAL : 0

Null Values in AMT_CREDIT : 0
Null Values in AMT_ANNUITY : 12
Null Values in AMT_GOODS_PRICE : 278

In [32]: app_data['AMT_ANNUITY'].fillna(app_data['AMT_ANNUITY'].median(),inplace=Tru
app_data['AMT_GOODS_PRICE'].fillna(app_data['AMT_GOODS_PRICE'].median(),inp
app_data['AMT_ANNUITY'].isnull().sum()
app_data['AMT_GOODS_PRICE'].isnull().sum()

Out[32]: 0

Correcting Data

In [33]: days = []
for i in app_data.columns:
if 'DAYS' in i:
days.append(i)
print('Unique Values in {0} column : {1}'.format(i,app_data[i].uniq
print('NULL Values in {0} column : {1}'.format(i,app_data[i].isnull
print()

Unique Values in DAYS_BIRTH column : [ -9461 -16765 -19046 ... -7951 -7

857 -25061]
NULL Values in DAYS_BIRTH column : 0

Unique Values in DAYS_EMPLOYED column : [ -637 -1188 -225 ... -12971

-11084 -8694]
NULL Values in DAYS_EMPLOYED column : 0

Unique Values in DAYS_REGISTRATION column : [ -3648. -1186. -4260. ...

-16396. -14558. -14798.]
NULL Values in DAYS_REGISTRATION column : 0

Unique Values in DAYS_ID_PUBLISH column : [-2120 -291 -2531 ... -6194 -5

854 -6211]
NULL Values in DAYS_ID_PUBLISH column : 0

Unique Values in DAYS_LAST_PHONE_CHANGE column : [-1134. -828. -815.

... -3988. -3899. -3538.]
NULL Values in DAYS_LAST_PHONE_CHANGE column : 1

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 13/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [34]: app_data[days]

Out[34]:
DAYS_BIRTH DAYS_EMPLOYED DAYS_REGISTRATION DAYS_ID_PUBLISH DAYS_LAST_PHO

0 -9461 -637 -3648.0 -2120

1 -16765 -1188 -1186.0 -291

2 -19046 -225 -4260.0 -2531

3 -19005 -3039 -9833.0 -2437

4 -19932 -3038 -4311.0 -3458

... ... ... ... ...

307506 -9327 -236 -8456.0 -1982

307507 -20775 365243 -4388.0 -4090

307508 -14966 -7921 -6737.0 -5150

307509 -11961 -4786 -2562.0 -931

307510 -16856 -1262 -5128.0 -410

307511 rows × 5 columns

In [35]: # Use absolute values in DAYS columns

app_data[days] = abs(app_data[days])
app_data[days]

Out[35]:
DAYS_BIRTH DAYS_EMPLOYED DAYS_REGISTRATION DAYS_ID_PUBLISH DAYS_LAST_PHO

0 9461 637 3648.0 2120

1 16765 1188 1186.0 291

2 19046 225 4260.0 2531

3 19005 3039 9833.0 2437

4 19932 3038 4311.0 3458

... ... ... ... ...

307506 9327 236 8456.0 1982

307507 20775 365243 4388.0 4090

307508 14966 7921 6737.0 5150

307509 11961 4786 2562.0 931

307510 16856 1262 5128.0 410

307511 rows × 5 columns

Binning

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 14/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [36]: # Lets do binning of these variables

for i in col:
app_data[i+'_Range']=pd.qcut(app_data[i],q=5,labels=['Very Low' , 'Low'
print(app_data[i+'_Range'].value_counts())
print()

Low 85756
High 75513
Very Low 63671
Very High 47118
Medium 35453
Name: AMT_INCOME_TOTAL_Range, dtype: int64

Very Low 64925

High 64024
Medium 61552
Very High 58912
Low 58098
Name: AMT_CREDIT_Range, dtype: int64

Medium 61569
Very Low 61507
Low 61499
Very High 61484
High 61452
Name: AMT_ANNUITY_Range, dtype: int64

Very Low 71454

Medium 61533
Very High 61430
High 61349
Low 51745
Name: AMT_GOODS_PRICE_Range, dtype: int64

In [37]: app_data['YEARS_EMPLOYED']= app_data['DAYS_EMPLOYED']/365

app_data['Client_Age']= app_data['DAYS_BIRTH']/365

In [38]: # Drop 'DAYS_EMPLOYED'& 'DAYS_BIRTH' column as we will be performing analys

app_data.drop(columns=['DAYS_EMPLOYED','DAYS_BIRTH'],inplace=True)

In [39]: app_data['Age Group']=pd.cut(

x=app_data['Client_Age'],
bins=[0,20,30,40,50,60,100],
labels=['0-20','20-30','30-40','40-50','50-60'
)

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 15/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [40]: app_data[['SK_ID_CURR','Client_Age','Age Group']]

Out[40]:
SK_ID_CURR Client_Age Age Group

0 100002 25.920548 20-30

1 100003 45.931507 40-50

2 100004 52.180822 50-60

3 100006 52.068493 50-60

4 100007 54.608219 50-60

... ... ... ...

307506 456251 25.553425 20-30

307507 456252 56.917808 50-60

307508 456253 41.002740 40-50

307509 456254 32.769863 30-40

307510 456255 46.180822 40-50

307511 rows × 3 columns

In [41]: app_data['Work Experience']=pd.cut(

x=app_data['YEARS_EMPLOYED'],
bins=[0,5,10,15,20,25,30,100],
labels=['0-5','5-10','10-15','15-20','20-25','
)

In [42]: app_data[['SK_ID_CURR','YEARS_EMPLOYED','Work Experience']]

Out[42]:
SK_ID_CURR YEARS_EMPLOYED Work Experience

0 100002 1.745205 0-5

1 100003 3.254795 0-5

2 100004 0.616438 0-5

3 100006 8.326027 5-10

4 100007 8.323288 5-10

... ... ... ...

307506 456251 0.646575 0-5

307507 456252 1000.665753 NaN

307508 456253 21.701370 20-25

307509 456254 13.112329 10-15

307510 456255 3.457534 0-5

307511 rows × 3 columns

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 16/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Outlier Detection

Analyzing AMT column for Outliers

In [43]: cols= ['AMT_INCOME_TOTAL','AMT_CREDIT', 'AMT_ANNUITY', 'AMT_GOODS_PRICE']

fig,axes = plt.subplots(ncols=2,nrows=2,figsize=(15,15))
count=0
for i in range(0,2):
for j in range(0,2):
sns.boxenplot(y=app_data[cols[count]],ax=axes[i,j])
count+=1
plt.show()

Below Columns have Outliers and those values can be dropped :-

AMT_INCOME_TOTAL
AMT_ANNUITY

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 17/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [44]: #Remove Outlier for 'AMT_INCOME_TOTAL' column

app_data=app_data[app_data['AMT_INCOME_TOTAL']<app_data['AMT_INCOME_TOTAL']
​
#Remove Outlier for 'AMT_ANNUITY' column
app_data=app_data[app_data['AMT_ANNUITY']<app_data['AMT_ANNUITY'].max()]

Analysing CNT_CHILDREN column for Outliers

In [45]: fig=px.box(app_data['CNT_CHILDREN'])
fig.update_layout(
title=dict(text = "Number of children",x=0.5,y=0.95),
title_font_size=20,
showlegend=False,
width =400,
height =400,
)
fig.show()

Number of children
20

15
value

10

0
CNT_CHILDREN

variable

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 18/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [46]: app_data['CNT_CHILDREN'].value_counts()

Out[46]: 0 215371
1 61118
2 26748
3 3717
4 429
5 84
6 21
7 7
14 3
8 2
9 2
12 2
10 2
19 2
11 1
Name: CNT_CHILDREN, dtype: int64

In [47]: app_data.shape[0]

Out[47]: 307509

In [48]: # Remove all data points where CNT_CHILDREN is greater than 10

app_data= app_data[app_data['CNT_CHILDREN']<=10]
app_data.shape[0]

Out[48]: 307501

Eight values dropped where number of children are greater than 10

Analysing YEARS_EMPLOYED column for Outliers

In [49]: sns.boxplot(y=app_data['YEARS_EMPLOYED'])

Out[49]: <AxesSubplot:ylabel='YEARS_EMPLOYED'>

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 19/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [50]: app_data['YEARS_EMPLOYED'].value_counts()

Out[50]: 1000.665753 55373

0.547945 156
0.613699 152
0.630137 151
0.545205 151
...
38.249315 1
32.402740 1
27.879452 1
25.915068 1
23.819178 1
Name: YEARS_EMPLOYED, Length: 12574, dtype: int64

In [51]: app_data.shape[0]

Out[51]: 307501

In [52]: app_data['YEARS_EMPLOYED'][app_data['YEARS_EMPLOYED']>1000]=np.NaN

In [53]: sns.boxplot(y=app_data['YEARS_EMPLOYED'])
plt.show()

In [54]: app_data.isnull().sum().sort_values(ascending=False).head(10)

Out[54]: Work Experience 55375

ORGANIZATION_TYPE 55373
YEARS_EMPLOYED 55373
NAME_TYPE_SUITE 1292
OBS_30_CNT_SOCIAL_CIRCLE 1021
DEF_30_CNT_SOCIAL_CIRCLE 1021
OBS_60_CNT_SOCIAL_CIRCLE 1021
DEF_60_CNT_SOCIAL_CIRCLE 1021
EXT_SOURCE_2 660
CNT_FAM_MEMBERS 2
dtype: int64

Analyzing AMT_REQ_CREDIT columns for Outliers

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 20/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [55]: cols = [i for i in app_data.columns if 'AMT_REQ' in i]

cols

Out[55]: ['AMT_REQ_CREDIT_BUREAU_HOUR',
'AMT_REQ_CREDIT_BUREAU_DAY',
'AMT_REQ_CREDIT_BUREAU_WEEK',
'AMT_REQ_CREDIT_BUREAU_MON',
'AMT_REQ_CREDIT_BUREAU_QRT',
'AMT_REQ_CREDIT_BUREAU_YEAR']

In [56]: fig,axes = plt.subplots(ncols=3,nrows=2,figsize=(15,15))

count=0
for i in range(0,2):
for j in range(0,3):
sns.boxenplot(y=app_data[cols[count]],ax=axes[i,j])
count+=1
plt.show()

AMT_REQ_CREDIT_BUREAU_QRT contains an outlier

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 21/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [57]: # Remove Outlier for AMT_REQ_CREDIT_BUREAU_QRT

app_data=app_data[app_data['AMT_REQ_CREDIT_BUREAU_QRT']<app_data['AMT_REQ_C

Univariate Analysis

In [58]: app_data.columns

Out[58]: Index(['SK_ID_CURR', 'TARGET', 'NAME_CONTRACT_TYPE', 'CODE_GENDER',

'CNT_CHILDREN', 'AMT_INCOME_TOTAL', 'AMT_CREDIT', 'AMT_ANNUITY',
'AMT_GOODS_PRICE', 'NAME_TYPE_SUITE', 'NAME_INCOME_TYPE',
'NAME_EDUCATION_TYPE', 'NAME_FAMILY_STATUS', 'NAME_HOUSING_TYPE',
'REGION_POPULATION_RELATIVE', 'DAYS_REGISTRATION', 'DAYS_ID_PUBLIS
H',
'OCCUPATION_TYPE', 'CNT_FAM_MEMBERS', 'REGION_RATING_CLIENT',
'REGION_RATING_CLIENT_W_CITY', 'WEEKDAY_APPR_PROCESS_START',
'HOUR_APPR_PROCESS_START', 'REG_REGION_NOT_LIVE_REGION',
'REG_REGION_NOT_WORK_REGION', 'LIVE_REGION_NOT_WORK_REGION',
'REG_CITY_NOT_LIVE_CITY', 'REG_CITY_NOT_WORK_CITY',
'LIVE_CITY_NOT_WORK_CITY', 'ORGANIZATION_TYPE', 'EXT_SOURCE_2',
'EXT_SOURCE_3', 'OBS_30_CNT_SOCIAL_CIRCLE', 'DEF_30_CNT_SOCIAL_CIR
CLE',
'OBS_60_CNT_SOCIAL_CIRCLE', 'DEF_60_CNT_SOCIAL_CIRCLE',
'DAYS_LAST_PHONE_CHANGE', 'AMT_REQ_CREDIT_BUREAU_HOUR',
'AMT_REQ_CREDIT_BUREAU_DAY', 'AMT_REQ_CREDIT_BUREAU_WEEK',
'AMT_REQ_CREDIT_BUREAU_MON', 'AMT_REQ_CREDIT_BUREAU_QRT',
'AMT_REQ_CREDIT_BUREAU_YEAR', 'AMT_INCOME_TOTAL_Range',
'AMT_CREDIT_Range', 'AMT_ANNUITY_Range', 'AMT_GOODS_PRICE_Range',
'YEARS_EMPLOYED', 'Client_Age', 'Age Group', 'Work Experience'],
dtype='object')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 22/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [59]: fig1=px.bar(app_data['OCCUPATION_TYPE'].value_counts(),color=app_data['OCCU
fig1.update_traces(textposition='outside',marker_coloraxis=None)
fig1.update_xaxes(title='Occupation Type')
fig1.update_yaxes(title='Count')
fig1.update_layout(
title=dict(text = "Occupation Type",x=0.5,y=0.95),
title_font_size=20,
showlegend=False,
height =450,
)
fig1.show()

Occupation Type

150k

100k
Count

50k

0
La Sa Co Ma Dr Hi Ac Me Se Co Cle Pr
bo les re na ive gh co dic cu ok an iv
re sta sta ger rs sk un i n rit ing i n
rs ff s ill tan es ys sta g
ff tec ts taf taf sta
hs f f ff
taf
f

Occupation Type

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 23/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [60]: fig2=px.bar(app_data['ORGANIZATION_TYPE'].value_counts(),color=app_data['OR
fig2.update_traces(textposition='outside',marker_coloraxis=None)
fig2.update_xaxes(title='Organization Type')
fig2.update_yaxes(title='Count')
fig2.update_layout(
title=dict(text = "Organization Type",x=0.5,y=0.95)
title_font_size=20,
showlegend=False,
height =450,
)
fig2.show()

Organization Type

60k
Count

40k

20k

0
Business Entity Type 3
Self-employed
Other
Medicine
Business Entity Type 2
Government
School
Trade: type 7
Kindergarten
Construction
Business Entity Type 1
Transport: type 4
Trade: type 3
Industry: type 9
Industry: type 3
Security
Housing
Industry: type 11
Military
Bank
Agriculture
Police
Transport: type 2
Postal
Security Ministries
Trade: type 2
Restaurant
Services
University
Industry: type 7
Transport: type 3
Industry: type 1
Hotel
Electricity
Industry: type 4
Trade: type 6
Industry: type 5
Insurance
Organization Type

Insights

Most People who applied for Loan application are Laborers

Most People who applied for Loan application belong to either Business Entity Type3 or Self-
Employed Organization Type.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 24/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [61]: cols = ['Age Group','NAME_CONTRACT_TYPE', 'NAME_INCOME_TYPE','NAME_EDUCATIO

'NAME_FAMILY_STATUS', 'NAME_HOUSING_TYPE','CODE_GENDER','Work Exper
​
#Subplot initialization
fig = make_subplots(
rows=4,
cols=2,
subplot_titles=cols,
horizontal_spacing=0.1,
vertical_spacing=0.13
)
# Adding subplots
count=0
for i in range(1,5):
for j in range(1,3):
fig.add_trace(go.Bar(x=app_data[cols[count]].value_counts().index,
y=app_data[cols[count]].value_counts(),
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (app_data[cols[co
),
row=i,col=j)
count+=1
fig.update_layout(
title=dict(text = "Analyze Categorical variables (Frequ
title_font_size=20,
showlegend=False,
width = 960,
height = 1600,
)
fig.show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 25/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 26/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Analyze Categorical variables (Frequency

Age Group N
80k 26.8% 90.5
24.9% 250k

22.1%
60k 200k

150k
40k 14.6%
11.6% 100k
20k
50k

0 0
30-40 40-50 50-60 20-30 60-100 0-20 Cash lo

NAME_INCOME_TYPE NA
51.6%

150k 71.0%
200k

150k
100k

100k
23.3%

50k
18.0%

2
50k

7.1%
0.0% 0.0% 0.0% 0.0%
0 0
Wo Co Pe St Un St Bu Ma Se
rk mm ns ate em ud sin ter co
ing er i o ne s p e n e nit nd
cia er loy t ss yl ar
r va ed m an ea y/
la nt ve
ss
oc
iat
e

NAME_FAMILY_STATUS N
200k
63.9% 88.7%
250k

150k
200k
localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 27/143
4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights 100k 150k

Bank has recieved majority of the loan application from 30-40 & 40-50 Age 100k groups.
50k 50% of clients who have applied for loan belong to Working Income Type.
More than
88.7% clients with Secondary/Secondary
14.8% Special education type have applied 50k for the loan.
6.4% 5.2%
Married people tend to apply more for loans. 63.9% clients who are have applied for loan are4.8%
9.7%

married.0 Ma Sin Civ Se Wi Un

0
Ho W
rri g il p a d o k n us
ed who lhave
Majority of the Clients e / applied ma for theraloan have
ted
w their own o house/apartment.
wn Around
e /a
n rri pa
88.7% clients are owning eitheroat m house or agan apartment.
e rt
ar
rie
Female loan applications are more as compared d to males. This may be because banks charge
less rate of interest for females.
Clients with work experience between 0-5 years have applied most for loan application.
90.5% Applicants have requested for Cash loans

CODE_GENDER
200k

54.1%
65.8%

150k 100k

100k 34.2%

25.7%
50k

50k

0 0
F M 0-5 5-10

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 28/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [62]: app_data.nunique().sort_values()

Out[62]: LIVE_REGION_NOT_WORK_REGION 2
REG_REGION_NOT_LIVE_REGION 2
REG_REGION_NOT_WORK_REGION 2
REG_CITY_NOT_LIVE_CITY 2
LIVE_CITY_NOT_WORK_CITY 2
REG_CITY_NOT_WORK_CITY 2
CODE_GENDER 2
NAME_CONTRACT_TYPE 2
TARGET 2
REGION_RATING_CLIENT 3
REGION_RATING_CLIENT_W_CITY 3
AMT_CREDIT_Range 5
AMT_INCOME_TOTAL_Range 5
AMT_REQ_CREDIT_BUREAU_HOUR 5
AMT_GOODS_PRICE_Range 5
NAME_EDUCATION_TYPE 5
Age Group 5
AMT_ANNUITY_Range 5
NAME_FAMILY_STATUS 6
NAME_HOUSING_TYPE 6
WEEKDAY_APPR_PROCESS_START 7
Work Experience 7
NAME_TYPE_SUITE 7
NAME_INCOME_TYPE 8
AMT_REQ_CREDIT_BUREAU_WEEK 9
AMT_REQ_CREDIT_BUREAU_DAY 9
DEF_60_CNT_SOCIAL_CIRCLE 9
AMT_REQ_CREDIT_BUREAU_QRT 10
DEF_30_CNT_SOCIAL_CIRCLE 10
CNT_CHILDREN 11
CNT_FAM_MEMBERS 12
OCCUPATION_TYPE 18
HOUR_APPR_PROCESS_START 24
AMT_REQ_CREDIT_BUREAU_MON 24
AMT_REQ_CREDIT_BUREAU_YEAR 25
OBS_60_CNT_SOCIAL_CIRCLE 33
OBS_30_CNT_SOCIAL_CIRCLE 33
ORGANIZATION_TYPE 57
REGION_POPULATION_RELATIVE 81
EXT_SOURCE_3 814
AMT_GOODS_PRICE 1002
AMT_INCOME_TOTAL 2547
DAYS_LAST_PHONE_CHANGE 3773
AMT_CREDIT 5603
DAYS_ID_PUBLISH 6168
YEARS_EMPLOYED 12573
AMT_ANNUITY 13671
DAYS_REGISTRATION 15688
Client_Age 17460
EXT_SOURCE_2 119830
SK_ID_CURR 307500
dtype: int64

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 29/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Checking Imbalance

In [63]: app_data['TARGET'].value_counts(normalize=True)

Out[63]: 0 0.919275
1 0.080725
Name: TARGET, dtype: float64

In [64]: fig=px.pie(values=app_data['TARGET'].value_counts(normalize=True),
names=app_data['TARGET'].value_counts(normalize=True).index,
hole = 0.5
)
fig.update_layout(
title=dict(text = "Target Imbalance",x=0.5,y=0.95),
title_font_size=20,
showlegend=False
)
fig.show()

Target Imbalance

8.07%

91.9%

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 30/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [65]: app_target0 = app_data.loc[app_data.TARGET == 0]

app_target1 = app_data.loc[app_data.TARGET == 1]

In [66]: app_target0.shape

Out[66]: (282677, 51)

In [67]: app_target1.shape

Out[67]: (24823, 51)

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 31/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [68]: ls = ['Age Group','NAME_CONTRACT_TYPE', 'NAME_INCOME_TYPE','NAME_EDUCATION_T

tle = [None]*(2*len(cols))
tle[::2]=[i+' (Non-Payment Difficulties)' for i in cols]
tle[1::2]=[i+' (Payment Difficulties)' for i in cols]

Subplot initialization
g = make_subplots(
rows=4,
cols=2,
subplot_titles=title,
)
Adding subplots
unt=0
or i in range(1,5):
for j in range(1,3):
if j==1:
fig.add_trace(go.Bar(x=app_target0[cols[count]].value_counts().ind
y=app_target0[cols[count]].value_counts(),
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (app_target0[cols[co
),
row=i,col=j)
else:
fig.add_trace(go.Bar(x=app_target1[cols[count]].value_counts().ind
y=app_target1[cols[count]].value_counts(),
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (app_target1[cols[co
),
row=i,col=j)
count+=1
g.update_layout(
title=dict(text = "Analyze Categorical variables (Payment/
title_font_size=20,
showlegend=False,
height = 1600,
)
g.show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 32/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 33/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Analyze Categorical variables (Payment/ Non-

Age Group (Non-Payment Difficulties) Age

8000
26.3% 31.8%
25.0%
60k 22.6%
6000
23

40k
14.1% 4000
12.0%

20k 2000

0 0
30-40 40-50 50-60 20-30 60-100 0-20 30-40 40

NAME_CONTRACT_TYPE (Non-Payment Difficulties) NAME_CON

250k
90.2% 93
20k
200k

15k
150k

10k
100k

50k 5k

9.8%
0 0
Cash loans Revolving loans Cash

NAME_INCOME_TYPE (Non-Payment Difficulties) NAME_INC

150k
15k
50.8%

61.3%

100k
10k

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 34/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

23.4%
50k 5k

18.5%
21

7.2%
0.0% 0.0% 0.0% 0.0%
0 0
Wo Co Pe St St Un Bu Ma Wo
rkin mm ns ate ud em sin ter r kin
er i o ne se e nt plo e nit
g cia rva ss yl g
r ye ma ea
la nt d n ve
ss
oc
iat
e

NAME_EDUCATION_TYPE (Non-Payment Difficulties) NAME_EDUC

200k 20k
70.3% 78.6%

150k 15k

100k 10k

25.1%
50k 5k

3.3%
1.2% 0.1%
0 0
Se Hi In Lo Ac Se
co gh co we ad co
nd er mp rs e mi nd
ar ed let ec cd ar
y /s uca eh on y
tio igh da eg
ec n ry ree
on er
da
ry
sp
ec
ial

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 35/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [69]: cols = ['NAME_FAMILY_STATUS', 'NAME_HOUSING_TYPE','CODE_GENDER','Work Exper

​
title = [None]*(2*len(cols))
title[::2]=[i+' (Non-Payment Difficulties)' for i in cols]
title[1::2]=[i+' (Payment Difficulties)' for i in cols]
​
#Subplot initialization
fig = make_subplots(
rows=4,
cols=2,
subplot_titles=title,
)
# Adding subplots
count=0
for i in range(1,5):
for j in range(1,3):
if j==1:
fig.add_trace(go.Bar(x=app_target0[cols[count]].value_counts().
y=app_target0[cols[count]].value_counts(),
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (app_target0[cols
),
row=i,col=j)
else:
fig.add_trace(go.Bar(x=app_target1[cols[count]].value_counts().
y=app_target1[cols[count]].value_counts(),
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (app_target1[cols
),
row=i,col=j)
count+=1
fig.update_layout(
title=dict(text = "Analyze Categorical variables (Payme
title_font_size=20,
showlegend=False,
height = 1600,
)
fig.show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 36/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 37/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Analyze Categorical variables (Payment/ Non-

NAME_FAMILY_STATUS (Non-Payment Difficulties) NAME_FAM

15k
64.2% 59.8%

150k

10k

100k

5k
50k
14.5%
5.4%
9.5%
6.4% 0.0%
0 0
Ma Sin Civ Se Wi Un Ma
rri gle il m pa do kn rri
ed /n ar ra w o wn ed
ria t e d
ot g
ma e
rri
ed

NAME_HOUSING_TYPE (Non-Payment Difficulties) NAME_HOU

250k
89.0% 20k 85.7%

200k
15k
150k

10k
100k

5k
50k
7
4.6% 3.6% 1.5% 0.9% 0.4%
0 0
Ho Wi Mu Re Of Co Ho
us th nic n ted fic -o us
e/ pa ipa ea pa e/
ap re la ap pa pa ap
ar nt pa ar rtm rtm
tm s r tm en
en tm en en
en t t t
t t

CODE_GENDER (Non-Payment Difficulties) CODE_

66.6% 57

150k
10k
localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 38/143
4/9/23, 6:41 PM EDA - Jupyter Notebook
10k

100k
33.4%
5k
50k

0 0
F M

Work Experience (Non-Payment Difficulties) Work Ex

15k
120k
52.9%

66.0%
100k

10k
80k

60k
26.1%

40k 5k

21.9%
20k
11.3%

4.5%
2.6% 1.4% 1.3%
0 0
0-5 5-10 10-15 15-20 20-25 25-30 30-100 0-5 5-1

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 39/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [70]: = ['OCCUPATION_TYPE', 'ORGANIZATION_TYPE' ,'AMT_INCOME_TOTAL_Range','AMT_CRE

= [None]*(2*len(cols))
[::2]=[i+' (Non-Payment Difficulties)' for i in cols]
[1::2]=[i+' (Payment Difficulties)' for i in cols]

lot initialization
make_subplots(
rows=4,
cols=2,
subplot_titles=title,
)
ing subplots
=0
in range(1,5):
or j in range(1,3):
if j==1:
fig.add_trace(go.Bar(x=app_target0[cols[count]].value_counts().index,
y=app_target0[cols[count]].value_counts(),
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (app_target0[cols[count
),
row=i,col=j)
else:
fig.add_trace(go.Bar(x=app_target1[cols[count]].value_counts().index,
y=app_target1[cols[count]].value_counts(),
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (app_target1[cols[count
),
row=i,col=j)
count+=1
pdate_layout(
title=dict(text = "Analyze Categorical variables (Payment/ No
title_font_size=20,
showlegend=False,
height = 1600,
)
how()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 40/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 41/143

 42/143
g
ORGANIZA
OCCUPAT

AMT_INCOME_TOTAL_Range (Non-Payment Difficulties) AMT_INCOME

Analyze Categorical variables (Payment/ Non-

1.7%
Trade: type 3
Core staff
2.2%
7.0% 2.2%
2.3%
2.4% Transport: type 4
Drivers

29.7%
8.5% 3.3%
3.4%
3.4% Medicine
12.5% Sales staff 3.6%
5.8%
4.1%
Business Entity Type 2
48.8% Laborers 17.9%
29.0%
Business Entity Type 3
12k

10k

8k

6k

4k

2k

6000

4000

2000

6000
ORGANIZATION_TYPE (Non-Payment Difficulties)
OCCUPATION_TYPE (Non-Payment Difficulties)

0.0%
0.2%

IT staff

0.1% 0.1% 0.1% 0.1% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
Industry: type 13
0.9% 0.6% 0.4% 0.4% 0.2% 0.2%

HR staff Industry: type 10

Realty agents Cleaning
Waiters/barmen staff Trade: type 1

0.3% 0.2% 0.2% 0.2% 0.2% 0.2% 0.2% 0.2% 0.2% 0.2%
EDA - Jupyter Notebook

Secretaries Culture
Low-skill Laborers Emergency
Private service staff Insurance

0.4% 0.4% 0.4% 0.3%

2.1% 1.9% 1.5%

Cleaning staff Electricity

20 7%
Cooking staff Transport: type 3

0.5% 0.5% 0.4%

Services

0.7% 0.6%
Security staff 0.8%
Medicine staff
0.8%
Security Ministries
2.8%

0.9%
0.9%
3.3%
Accountants
1.0%
1.0%
Agriculture

24.6%
1.0%
1.1%
Industry: type 11
High skill tech staff
3.8%
1.1%
1.2%
1.3%
Security

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb
5.8% Drivers
1.3%
1.4%
1.4%
Industry: type 9
Managers
2.1%
7.1% 2.4%
2.6%
2.8% Construction
Core staff

27.7%
9.1% 3.1%
3.6%
4.2% School
10.3% Sales staff 4.2%
6.7%
4.5%
Medicine
49.3% Laborers 15.0%
26.8%
Business Entity Type 3
100k

50k

60k

50k

40k

30k

20k

10k

80k

60k
4/9/23, 6:41 PM
4/9/23, 6:41 PM EDA - Jupyter Notebook
20.7%

4000
40k 15.6%
Bivariate / Multivariate Analysis
11.5%
In [71]: # Group 20k
data by 'AMT_CREDIT_Range' & 'CODE_GENDER' 2000
df1=app_data.groupby(by=['AMT_CREDIT_Range','CODE_GENDER']).count().reset_i
df1
0 0
Out[71]: Low High Very Low Very High Medium Low
AMT_CREDIT_Range CODE_GENDER SK_ID_CURR

0 Very Low F 44073

1 Very Low M 20850

2 Low F 38185

3 Low M 19912

4 Medium F 39807
AMT_CREDIT_Range (Non-Payment Difficulties) AMT_CRE
5 60k Medium M 21743
21.3% 6000 24.9%
20.9%
6 High F 19.6% 42360
19.6%
50k 18.7% 5000
7 High M 21662
40k 4000
8 Very High F 38019

9 30k Very High M 20889 3000

20k 2000
In [72]: # Group data by 'AMT_INCOME_TOTAL_Range' & 'CODE_GENDER'
df2=app_data.groupby(by=['AMT_INCOME_TOTAL_Range','CODE_GENDER']).count().r
10k 1000
df2

Out[72]: 0 0
Very Low High Medium Very High Low Medium
AMT_INCOME_TOTAL_Range CODE_GENDER SK_ID_CURR

0 Very Low F 51464

1 Very Low M 12205

2 Low F 59967

3 Low M 25787

4 Medium F 23184

5 Medium M 12269

6 High F 43441

7 High M 32070

8 Very High F 24388

9 Very High M 22725

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 43/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [73]: fig1=px.bar(data_frame=df1,
x='AMT_CREDIT_Range',
y='SK_ID_CURR',color='CODE_GENDER',
barmode='group',
text='SK_ID_CURR'
)
fig1.update_traces(textposition='outside')
fig1.update_xaxes(title='Day')
fig1.update_yaxes(title='Transaction count')
fig1.update_layout(
title=dict(text = "Loan Applications by Gender & Credit
title_font_size=20,
)
fig1.show()
​

Loan Applications by Gender & Cre

45k 44073
42360
39807
40k 38185

35k

30k
Transaction count

25k
21743
20850
19912
20k

15k

10k

5k

0
Very Low Low Medium Hig

Day

Insights

Females are mostly applying for Very Low credit loans.

Males are applying for Medium & High credit loans.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 44/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [74]: fig2=px.bar(data_frame=df2,
x='AMT_INCOME_TOTAL_Range',
y='SK_ID_CURR',color='CODE_GENDER',
barmode='group',
text='SK_ID_CURR'
)
fig2.update_traces(textposition='outside')
fig2.update_xaxes(title='Day')
fig2.update_yaxes(title='Transaction count')
fig2.update_layout(
title=dict(text = "Loan Applications by Gender & Total
title_font_size=20,
)
fig2.show()

Loan Applications by Gender & Total In

59967
60k

51464
50k

43441
Transaction count

40k

30k
25787
23184

20k

12205 12269
10k

0
Very Low Low Medium Hig

Day

Insights

Females with Low & Very Low total income have applied the most for the loan.

Education Type VS Credit Amount (Payment / Non Payment Difficulties)

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 45/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [75]: fig = px.box(app_target0, x="NAME_EDUCATION_TYPE", y="AMT_CREDIT", color='N

title="Education Type VS Credit Amount (Non Payment Difficulti
fig.show()

Education Type VS Credit Amount (Non Payment Difficulties)

4M

3M
AMT_CREDIT

2M

1M

0
Hi Se In Lo
gh co co we
er nd mp rs
ed ar let e
uc y/ eh
atio se igh
n co er
nd
ar
ys
pec
ial

NAME_EDUCATION_TYPE

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 46/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [76]: fig = px.box(app_target1, x="NAME_EDUCATION_TYPE", y="AMT_CREDIT", color='N

title="Education Type VS Credit Amount (Payment Difficulties)"
fig.show()

Education Type VS Credit Amount (Payment Difficulties)

4M

3M
AMT_CREDIT

2M

1M

0
Se In Hi Lo
co co gh we
nd mp er rs
ar let ed e
y/ e uc
se hig atio
co he n
nd r
ar
y sp
ec
ial

NAME_EDUCATION_TYPE

Income VS Credit Amount (Payment / Non Payment Difficulties)

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 47/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [77]: fig = px.box(app_target0, x="AMT_INCOME_TOTAL_Range", y="AMT_CREDIT", color

title="Income Range VS Credit Amount (Non-Payment Difficulties
fig.show()

Income Range VS Credit Amount (Non-Payment Difficulties)

4M

3.5M

3M
AMT_CREDIT

2.5M

2M

1.5M

1M

0.5M

0
Very High Very Low High Low

AMT_INCOME_TOTAL_Range

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 48/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [78]: fig = px.box(app_target1, x="AMT_INCOME_TOTAL_Range", y="AMT_CREDIT", color

title="Income Range VS Credit Amount (Payment Difficulties)")
fig.show()

Income Range VS Credit Amount (Payment Difficulties)

4M

3.5M

3M
AMT_CREDIT

2.5M

2M

1.5M

1M

0.5M

0
High Very High Medium Low

AMT_INCOME_TOTAL_Range

Age Group VS Credit Amount (Payment / Non Payment Difficulties)

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 49/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [79]: fig = px.box(app_target0, x="Age Group", y="AMT_CREDIT", color='NAME_FAMILY

title="Age Group VS Credit Amount (Non-Payment Difficulties)")
fig.show()

Age Group VS Credit Amount (Non-Payment Difficulties)

4M

3.5M

3M
AMT_CREDIT

2.5M

2M

1.5M

1M

0.5M

0
40-50 30-40 50-60 20-30

Age Group

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 50/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [80]: fig = px.box(app_target1, x="Age Group", y="AMT_CREDIT", color='NAME_FAMILY

title="Age Group VS Credit Amount (Payment Difficulties)")
fig.show()

Age Group VS Credit Amount (Payment Difficulties)

4M

3.5M

3M
AMT_CREDIT

2.5M

2M

1.5M

1M

0.5M

0
20-30 40-50 30-40 50-60

Age Group

Work Experience VS Credit Amount (Payment / Non Payment Difficulties)

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 51/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [81]: fig = px.box(app_target0, x="Work Experience", y="AMT_CREDIT", color='NAME_

title="Work Experience VS Credit Amount (Non-Payment Difficult
fig.show()

Work Experience VS Credit Amount (Non-Payment Difficulties)

4M

3.5M

3M
AMT_CREDIT

2.5M

2M

1.5M

1M

0.5M

0
0-5 5-10 10-15 25-30 15-20 2

Work Experience

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 52/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [82]: fig = px.box(app_target1, x="Work Experience", y="AMT_CREDIT", color='NAME_

title="Work Experience VS Credit Amount (Payment Difficulties)
fig.show()

Work Experience VS Credit Amount (Payment Difficulties)

4M

3.5M

3M
AMT_CREDIT

2.5M

2M

1.5M

1M

0.5M

0
0-5 5-10 20-25 10-15 15-20 2

Work Experience

Numerical vs Numerical Variables

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 53/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [83]: sns.pairplot(app_data[['AMT_INCOME_TOTAL', 'AMT_GOODS_PRICE',

'AMT_CREDIT', 'AMT_ANNUITY',
'Client_Age','YEARS_EMPLOYED' ]].fillna(0))
plt.show()

Correlation in target0 & target1

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 54/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [84]: plt.figure(figsize=(12,8))
sns.heatmap(app_target0[['AMT_INCOME_TOTAL', 'AMT_GOODS_PRICE',
'AMT_CREDIT', 'AMT_ANNUITY',
'Client_Age','YEARS_EMPLOYED' ,
'DAYS_ID_PUBLISH', 'DAYS_REGISTRATION',
'EXT_SOURCE_2','EXT_SOURCE_3','REGION_POPULATION_REL
plt.title('Correlation matrix for Non-Payment Difficulties')
plt.show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 55/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [85]: plt.figure(figsize=(12,8))
sns.heatmap(app_target1[['AMT_INCOME_TOTAL', 'AMT_GOODS_PRICE',
'AMT_CREDIT', 'AMT_ANNUITY',
'Client_Age','YEARS_EMPLOYED' ,
'DAYS_ID_PUBLISH', 'DAYS_REGISTRATION',
'EXT_SOURCE_2','EXT_SOURCE_3','REGION_POPULATION_REL
plt.title('Correlation Matrix for Payment Difficulties')
plt.show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 56/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Data Analysis on Previous Application dataset

In [86]: appdata_previous = pd.read_csv('previous_application.csv')

appdata_previous.head()

Out[86]:
SK_ID_PREV SK_ID_CURR NAME_CONTRACT_TYPE AMT_ANNUITY AMT_APPLICATION AMT_CR

0 2030495 271877 Consumer loans 1730.430 17145.0 17

1 2802425 108129 Cash loans 25188.615 607500.0 679

2 2523466 122040 Cash loans 15060.735 112500.0 136

3 2819243 176158 Cash loans 47041.335 450000.0 470

4 1784265 202054 Cash loans 31924.395 337500.0 404

5 rows × 37 columns

Drop Columns with NULL Values greater than 40%

In [87]: s1= (appdata_previous.isnull().mean()*100).sort_values(ascending=False)[app

s1

Out[87]: RATE_INTEREST_PRIVILEGED 99.643698

RATE_INTEREST_PRIMARY 99.643698
AMT_DOWN_PAYMENT 53.636480
RATE_DOWN_PAYMENT 53.636480
NAME_TYPE_SUITE 49.119754
NFLAG_INSURED_ON_APPROVAL 40.298129
DAYS_TERMINATION 40.298129
DAYS_LAST_DUE 40.298129
DAYS_LAST_DUE_1ST_VERSION 40.298129
DAYS_FIRST_DUE 40.298129
DAYS_FIRST_DRAWING 40.298129
dtype: float64

In [88]: appdata_previous.shape

Out[88]: (1670214, 37)

In [89]: appdata_previous.drop(columns = s1.index,inplace=True)

In [90]: appdata_previous.shape

Out[90]: (1670214, 26)

Changing negative values in the DAYS columns to positive values

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 57/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [91]: days = []
for i in appdata_previous.columns:
if 'DAYS' in i:
days.append(i)
print('Unique Values in {0} column : {1}'.format(i,appdata_previous
print()

Unique Values in DAYS_DECISION column : [ -73 -164 -301 ... -1967 -238
9 -1]

In [92]: appdata_previous[days]= abs(appdata_previous[days])

In [93]: appdata_previous[days]

0 73

1 164

2 301

3 512

4 781

... ...

1670209 544

1670210 1694

1670211 1488

1670212 1185

1670213 1193

1670214 rows × 1 columns

In [94]: # Replcae XNA and XAP are replaced by NaN

appdata_previous=appdata_previous.replace('XNA', np.NaN)
appdata_previous=appdata_previous.replace('XAP', np.NaN)

Univariate Analysis on Previous Application Data

In [95]: appdata_previous.columns

Out[95]: Index(['SK_ID_PREV', 'SK_ID_CURR', 'NAME_CONTRACT_TYPE', 'AMT_ANNUITY',

'AMT_APPLICATION', 'AMT_CREDIT', 'AMT_GOODS_PRICE',
'WEEKDAY_APPR_PROCESS_START', 'HOUR_APPR_PROCESS_START',
'FLAG_LAST_APPL_PER_CONTRACT', 'NFLAG_LAST_APPL_IN_DAY',
'NAME_CASH_LOAN_PURPOSE', 'NAME_CONTRACT_STATUS', 'DAYS_DECISION',
'NAME_PAYMENT_TYPE', 'CODE_REJECT_REASON', 'NAME_CLIENT_TYPE',
'NAME_GOODS_CATEGORY', 'NAME_PORTFOLIO', 'NAME_PRODUCT_TYPE',
'CHANNEL_TYPE', 'SELLERPLACE_AREA', 'NAME_SELLER_INDUSTRY',
'CNT_PAYMENT', 'NAME_YIELD_GROUP', 'PRODUCT_COMBINATION'],
dtype='object')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 58/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [96]: cols = ['NAME_CONTRACT_STATUS','WEEKDAY_APPR_PROCESS_START',

'NAME_PAYMENT_TYPE','CODE_REJECT_REASON',
'NAME_CONTRACT_TYPE','NAME_CLIENT_TYPE']
​
#Subplot initialization
fig = make_subplots(
rows=3,
cols=2,
subplot_titles=cols,
horizontal_spacing=0.1,
vertical_spacing=0.17
)
# Adding subplots
count=0
for i in range(1,4):
for j in range(1,3):
fig.add_trace(go.Bar(x=appdata_previous[cols[count]].value_counts()
y=appdata_previous[cols[count]].value_counts()
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (appdata_previous
),
row=i,col=j)
count+=1
fig.update_layout(
title=dict(text = "Analyze Categorical variables (Frequ
title_font_size=20,
showlegend=False,
width = 960,
height = 1200,
)
fig.show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 59/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 60/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Analyze Categorical variables (Frequency

NAME_CONTRACT_STATUS WEEKDAY_APPR_
1M 250k

15.3%

15.3%
62.1%

15.2%

15.1%
0.8M 200k

0.6M 150k

0.4M 100k

18.9% 17.4%
0.2M 50k

1.6%
0 0
Approved Canceled Refused Unused offer TU WE MO
ES DN ND
DA ES AY
Y DA
Y

NAME_PAYMENT_TYPE CODE_REJE
1M

56.2%
99.1%
150k
0.8M

0.6M 100k

0.4M
50k
17.9%

0.2M
12.0%

8.5
0.8% 0.1%
0 0
Ca No Ca HC LIM SC C
sh n- sh IT O
th ca les
ro sh sf
ug fro ro
ht m m
he yo th
ba u ea
n ra cc
k c co oun
un to
t ft
h ee
mp
loy
er

NAME_CONTRACT_TYPE NAME_CLI
44.8% 1.2M 73.8%
43.7%

600k 1M

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 61/143

4/9/23, 6:41 PM EDA - Jupyter Notebook
0.8M

400k
Approved Loans 0.6M

0.4M
In [97]: approved=appdata_previous[appdata_previous['NAME_CONTRACT_STATUS']=='Approv
200k
11.6% 18.
0.2M

0 0
Cash loans Consumer loans Revolving loans Repeater Ne

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 62/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [98]:s = ['NAME_PORTFOLIO','NAME_GOODS_CATEGORY',
'CHANNEL_TYPE','NAME_YIELD_GROUP' , 'NAME_PRODUCT_TYPE','NAME_CASH_LOAN_

bplot initialization
= make_subplots(
rows=3,
cols=2,
subplot_titles=cols,
horizontal_spacing=0.1,
vertical_spacing=0.19
)
dding subplots
nt=0
i in range(1,4):
for j in range(1,3):
fig.add_trace(go.Bar(x=approved[cols[count]].value_counts().index,
y=approved[cols[count]].value_counts(),
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (approved[cols[count]
),
row=i,col=j)
count+=1
update_layout(
title=dict(text = "Analyze Categorical variables (Frequency
title_font_size=20,
showlegend=False,
width = 960,
height = 1400,
)
show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 63/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 64/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Analyze Categorical variables (Frequency

NAME_PORTFOLIO NAM
600k

29.7%
60.4%

150k

400k

17.8%
100k
30.1%

14.3%

14.1%
200k
50k

7.8%
9.4%

3.6%

3.4%
0.0%
0 0
POS Cash Cards Cars

Mobile

Audio/Video

Furniture

Clothing and Accessories

CHANNEL_TYPE N
400k
38.8%

300k 34.4%

300k
27.9%

200k

200k
18.2%

100k
100k
9.3%

3.0% 2.4%
0.3% 0.0%
0 0
Co Cr St Re AP Co Ch Ca middle
u nt e dit on g ion + nt an rd
ry- e ( Ca a ct n el ea
an al s c ler
wi dc /L hl en of
de a oc oa ter co
sh al n) rp
off or
ic ate
es sa
les

NAME_PRODUCT_TYPE NAME
3

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 65/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

32.2%
82.3% 8000
300k

Refused Loans

24.8%
6000
200k
In [99]: refused=appdata_previous[appdata_previous['NAME_CONTRACT_STATUS']=='Refused
4000

13.3%
100k
2000
17.7%

4.6%

3.9%

3.3%
0 0
x-sell walk-in

Repairs
Other
Urgent needs
Everyday expenses
Medicine
Buying a used car
Education

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 66/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [100]: cols = ['NAME_PORTFOLIO','NAME_GOODS_CATEGORY',

'CHANNEL_TYPE','NAME_YIELD_GROUP' , 'NAME_PRODUCT_TYPE','NAME_CASH_
​
#Subplot initialization
fig = make_subplots(
rows=3,
cols=2,
subplot_titles=cols,
horizontal_spacing=0.1,
vertical_spacing=0.19
)
# Adding subplots
count=0
for i in range(1,4):
for j in range(1,3):
fig.add_trace(go.Bar(x=refused[cols[count]].value_counts().index,
y=refused[cols[count]].value_counts(),
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (refused[cols[cou
),
row=i,col=j)
count+=1
fig.update_layout(
title=dict(text = "Analyze Categorical variables (Frequ
title_font_size=20,
showlegend=False,
width = 960,
height = 1400,
)
fig.show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 67/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 68/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Analyze Categorical variables (Frequency

NAME_PORTFOLIO NAM
20k

30.6%
55.7%

100k 15k

20.2%
10k
25.5%

13.6%

13.6%
50k
18.7%
5k

6.5%

3.7%

3.4%
0.0%
0 0
Cash POS Cards Cars

Mobile

Consumer Electronics

Furniture

Photo / Cinema Equipment

CHANNEL_TYPE N
150k
51.8%

34.7%
60k

100k
40k
22.6%

50k
20k

7.6% 7.6% 3.9%

5.4% 1.2% 0.0%
0 0
Cr Co AP St Co Re Ch Ca low_normal
e dit u nt + on nt g ion an rd
ry- (C e ac n el ea
an as tc al ler
dc wi hl en /L of
as de oa ter oc co
ho n) al rp
ffic or
ate
es sa
les

NAME_PRODUCT_TYPE NAME
15k
3

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 69/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

35.3%
59.3%
100k

Merging
80k Application & Previous Application Data
10k
40.7%
60k

20.9%
In [101]: appdata_merge = app_data.merge(appdata_previous,on='SK_ID_CURR', how='inner
appdata_merge.shape
40k 5k
Out[101]: (1413670, 76)

11.5%
20k
Analysis of Merged Data

4.7%

4.6%

3.8%
0 0
x-sell walk-in

Repairs
Other
Urgent needs
Building a house or an annex
Buying a used car
Payments on other loans
Everyday expenses
In [102]: # Function for multiple plotting - Bar Chart
def plot_merge(appdata_merge,column_name):
col_value = ['Refused','Approved', 'Canceled' , 'Unused offer']

#Subplot initialization
fig = make_subplots(
rows=2,
cols=2,
subplot_titles=col_value,
horizontal_spacing=0.1,
vertical_spacing=0.3
)
# Adding subplots
count=0
for i in range(1,3):
for j in range(1,3):
fig.add_trace(go.Bar(x=appdata_merge[appdata_merge['NAME_CONTRAC
y=appdata_merge[appdata_merge['NAME_CONTRACT_ST
name=cols[count],
textposition='auto',
text= [str(i) + '%' for i in (appdata_merge[app
),
row=i,col=j)
count+=1
fig.update_layout(
title=dict(text = "NAME_CONTRACT_STATUS VS "+column_name
title_font_size=20,
showlegend=False,
width = 960,
height = 960,
)
fig.show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 70/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [103]: # Function for multiple plotting - Pie Chart

def plot_pie_merge(appdata_merge,column_name):
col_value = ['Refused','Approved', 'Canceled' , 'Unused offer']

#Subplot initialization
fig = make_subplots(
rows=2,
cols=2,
subplot_titles=col_value,
specs=[[{"type": "pie"}, {"type": "pie"}],[{"type": "p
)
# Adding subplots
count=0
for i in range(1,3):
for j in range(1,3):
fig.add_trace(go.Pie(labels=appdata_merge[appdata_merge['NAME_C
values=appdata_merge[appdata_merge['NAME_CONTR
textinfo='percent',
insidetextorientation='auto',
hole=.3
),
row=i,col=j)
count+=1
fig.update_layout(
title=dict(text = "NAME_CONTRACT_STATUS VS "+column_nam
title_font_size=20,
width = 960,
height = 960,
)
fig.show()

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 71/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [104]: plot_pie_merge(appdata_merge,'NAME_CONTRACT_TYPE_y')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 72/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 73/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS NAME_CON

Refused Ap

26.2%
30.1%

56.9%

16.9%
9.3%

1.9%
Canceled Unus
0.0176%

0.513% 14.4%

85%

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 74/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Banks mostly approve Consumer Loans

Most of the _Refused & Cancelled loans are cash loans.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 75/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [105]: plot_pie_merge(appdata_merge,'NAME_CLIENT_TYPE')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 76/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 77/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS NAME_C

Refused

5.11%
5.0
4%
27.4%

9.36%

89.8%

Canceled U
1.13%
6.23%

15

8.21%

92.6%

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 78/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Most of the approved , refused & canceled loans belong to the old clients.
Almost 27.4% loans were provided to new customers.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 79/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [106]: plot_pie_merge(appdata_merge,'CODE_GENDER')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 80/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 81/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS CODE

Refused

33.4% 32.

66.6%

Canceled

31%
37.8

69%

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 82/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Approved percentage of loans provided to females is more as compared to refused

percentage.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 83/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [107]: plot_merge(appdata_merge,'NAME_EDUCATION_TYPE')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 84/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 85/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS NAME_EDU

Refused

73.2% 73.6%
600k
150k
500k

400k
100k
300k

200k
50k 22.0% 2

100k
3.5%
1.2% 0.0%
0 0
Se Hi In Lo Ac Se
co gh c om we ad co
n da er ple rs em nda
r ed t e co ic ry
y/ uc eh nd de /
se ati igh ar gr
c on on er y ee
da
ry
sp
ec
ial

Canceled
200k
15k
73.7% 65.4%

150k

10k

100k

2
5k
50k 22.0%

3.1%
1.2% 0.0%
0 0
Se Hi In Lo Ac Se
co gh co we ad co
nd er mp rs em nd
ar ed let ec ic ar
y/ uc e on de y/
se ati hig d ar gr
co on h er y ee
nd
ar
ys
p ec
ial

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 86/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Most of the approved loans belong to applicants with Secondary / Secondary Special
education type.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 87/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [108]: plot_merge(appdata_merge,'NAME_INCOME_TYPE')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 88/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 89/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS NAME_IN

Refused

52.1%

51.4%
120k
400k

100k

300k
80k

60k 23.7% 200k

22.4%
40k
17.9%
100k
20k
6.4%
0.0% 0.0%
0 0
Wo Co Pe St Un Ma St Wo Co
rk mm ns ate em ter ud r kin m
ing er ion s p n en
cia er e rva loy ity t g
la nt e d lea
ss ve
oc
iat
e

Canceled
14k
48.7%

120k 62.6%

12k
100k
10k
80k
8k
60k
23.0%

6k
22.2%

40k 22.3
4k

20k 2k
6.0%
0.0% 0.0%
0 0
Wo Co Pe St Un St Ma Wo Co
rk m n sio ate em ud ter rk
in me ne se p e n nit in
g rci r rva loy t yl g
al nt e d ea
as ve
so
cia
te

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 90/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Across all Contract Status (Approved , Refused , Canceled , Unused Offer) people with
Working income type are leading. So it is quite evident that majority of the loans are coming
from this income type class.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 91/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [109]: plot_pie_merge(appdata_merge,'NAME_FAMILY_STATUS')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 92/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 93/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS NAME_FA

Refused Appro

14% 13%

9.8%
11.5%

6.4%
6.71%
62%
5.7%
5.81%

Canceled Unused

12.9%
17.9%

10.1%

8.61%

6.37%

64.5% 7.03%
6.22%
%
32
3.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 94/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Approved percentage of loans for married applicants is higher than the rest of the contract
status (refused , canceled etc.).

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 95/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [110]: plot_pie_merge(appdata_merge,'NAME_PORTFOLIO')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 96/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 97/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS NAME_

Refused

25.8%
30.1%

55.5%

18.6%
9.3%

0.0272%

0.0503%

1.45% Canceled
0.377%
4.76%

93.4%

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 98/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

60.6% previous approved loans belong to POS name portfolio.

Majority of the loans refused were cash loans.
93.4% loans that belong to POS were canceled

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 99/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [111]: plot_merge(appdata_merge,'OCCUPATION_TYPE')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 100/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 101/143

 102/143
NAME_CONTRACT_STATUS VS OCCUPA

5.9% Drivers 6.8% Drivers

6.6% Managers 7.0% Managers
8.5% Core staff 10.5% Core staff
10.5% Sales staff 11.5% Sales staff
50.2% Laborers 44.3% Laborers
400k

300k

200k

100k

10k

8k

6k

4k

2k

0
0.2% 0.1%

0.3% 0.2% 0.1%

IT staff IT staff
HR staff HR staff
0.8% 0.7% 0.4% 0.3% 0.2%

Realty agents Realty agents

0.9% 0.7% 0.4% 0.4%

Secretaries Secretaries
EDA - Jupyter Notebook

Waiters/barmen staff Waiters/barmen staff

Low-skill Laborers Low-skill Laborers
Private service staff Private service staff
2.1% 1.8%

2.3% 2.0% 1.8%

Canceled
Cleaning staff Cleaning staff
Refused

Cooking staff Cooking staff

2.3%
Security staff Security staff
Accountants Medicine staff
2.8%

2.6%
Medicine staff Accountants

2.8%

2.9%
High skill tech staff High skill tech staff

3.2%

3.3%

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb
6.3% Drivers 5.7% Drivers
7.4% Managers 6.5% Managers
8.4% Core staff 7.7% Core staff
11.4% Sales staff 10.7% Sales staff
48.9% Laborers 51.6% Laborers
120k

100k

80k

60k

40k

20k

140k

120k

100k

80k

60k

40k

20k

0
4/9/23, 6:41 PM
4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 103/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [112]: plot_merge(appdata_merge,'NAME_GOODS_CATEGORY')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 104/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 105/143

 106/143
NAME_CONTRACT_STATUS VS NAME_GOO

3.4%
Photo / Cinema Equipment Consumer Electronics

3.6%
3.6%
Audio/Video

3.7%
7.7%
Furniture
13.9% 5.3%
Photo / Cinema Equipment
14.3%
Audio/Video
17.9%
15.9% Computers
29.9%
Mobile 69.2% Mobile
150k

100k

50k

15k

10k

5k

0
sm
mp udio ons urni hoto ons uto edi ewe ouri

a E Mat ries lies

Fitness
0.0% 0.0% 0.0% 0.0% 0.0%

1.2% 0.6% 0.6% 0.6%

T
ry
Weapon

p
tio ces l Sup
l
Insurance

me ls
uip ria
nt
Additional Service

so
c a

e
Education
Direct Sales
0.4% 0.3% 0.2% 0.2% 0.1% 0.1%
EDA - Jupyter Notebook

M
A c
em n
q
Medicine
Other

/ C ruc
A
Tourism

t
n
Gardening

ics
3.1%

C
Canceled
Office Appliances
Refused

on
Sport and Leisure
0.6% 0.5% 0.4% 0.4%

ctr
er Vide mer ure
4.4%
Medical Supplies

Ele
P
t
Vehicles
Homewares 5.0%

F
u
Jewelry
1.0% 0.9%

o
Auto Accessories
13.1%

C
Clothing and Accessories

/
2.9%

s
Photo / Cinema Equipment

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb
3.5%
15.0%

ut
A
3.7%
Construction Materials
6.4%
Furniture

Mo Co
13.5%
Consumer Electronics 16.2%

e
Audio/Video

bil
13.6%
20.2%
Computers 40.0%
30.9%
Mobile
15k

10k

5k

60

50

40

30

20

10

0
4/9/23, 6:41 PM
4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 107/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [113]: plot_merge(appdata_merge,'PRODUCT_COMBINATION')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 108/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 109/143

 110/143
8.1% Cash X-Sell: low
NAME_CONTRACT_STATUS VS PRODUCT_

S
ew
12.7%

PO
8.5% POS industry with interest

bil
mo
11.1% Cash X-Sell: middle

S
POS mobile with interest

PO
17.0%
81.3%
22.6% POS household with interest
200k

150k

100k

50k

15k

10k

5k

0
0.1%

0.1% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0%

POS others without interest POS mobile without interest
0.5% 0.4%

POS industry without interest POS household without interest

POS mobile without interest
POS industry with interest
POS other with interest
0.8%
EDA - Jupyter Notebook

Card X-Sell
3.0% POS industry with interest
POS household with interest
3.5% POS household without interest
4.2% Cash X-Sell: high Cash Street: high

Canceled
Refused

5.8% Cash Street: middle POS mobile with interest

1.5% 0.8% 0.4% 0.3% 0.2% 0.2%

7.2% Card X-Sell Cash Street: middle
7.3% Cash Street: high Cash X-Sell: high
7.9% Cash Street: low POS other with interest
8.1% Cash X-Sell: middle
Cash Street: low
8.6% POS mobile with interest

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb
Cash X-Sell: middle
9.2% Cash
Cash X-Sell: low
9.4% POS household with interest
9.7% Card Street
14.4% Card Street
14.4% Cash X-Sell: low 81.9% Cash
30k

20k

10k

200k

150k

100k

50k

0
4/9/23, 6:41 PM
4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Most of the approved loans belong to POS hosehold with interest & POS mobile with
interest product combination.
15% refused loans belong to Cash X-Sell: low product combination.
Most of the canceled loans belong to Cash category.
81.3% Unused Offer loans belong to POS mobile with interest.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 111/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [114]: plot_merge(appdata_merge,'NAME_PAYMENT_TYPE')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 112/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 113/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS NAME_PA

Refused Appr

150k 99.2% 99.0%

600k

100k
400k

50k
200k

0.6% 0.1% 0.9

0 0
Ca No Ca Ca
sh n- sh sh
th ca les th
r ou sh sf ro
gh fro ro ug
m m ht
th yo th he
eb u ea ba
a nk ra cc
cc ou
ou n to
n t f th
ee
mp
loy
er

Canceled Unuse

2500 98.3% 99.8%

20k

2000
15k

1500

10k
1000

5k
500

1.2% 0.6% 0.1

0 0
Ca No Ca Ca
sh n- sh sh
th ca les th
ro sh sf ro
u gh fro ro u gh
m m
th yo th th
eb ur ea eb
an ac c co a
k c un
ou to
nt
f th
ee
mp
loy
er

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 114/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 115/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [115]: plot_merge(appdata_merge,'CHANNEL_TYPE')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 116/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 117/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS CHAN

Refused
350k

51.2%

38.9%
120k
300k
100k
250k

27.6%
80k
200k

60k
150k
22.9%

40k
100k

20k 50k
3.9%
7.9%

7.7%
5.2% 1.2% 0.0%
0 0
Cr Co AP St Co Re Ch Ca Co Cr
ed un + on nt gio an rd un ed
it try (C e ac n ne ea try it
an - a s t c a l l ler -w
dc w ide hl en /L o fc ide
as oa ter oc o r
ho n) al po
ffic ra
es te
sa
les

Canceled
88.6%

20k 92.0%
200k

15k
150k

10k
100k

50k 5k

9.4% 3
1.2% 0.6% 0.0% 0.0% 0.0% 0.0%
0 0
Cr Co AP Co Ch Re St Ca Co
e dit ntac + un an gio on rd un
( Ca t ry- n el n e ea try
an tc s al ler -w
dc en hl w i o f / i
as ter oa de co L oc de
ho n) rp a l
ffic or
ate
es sa
les

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 118/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Most of the approved loans belong to either Country-wide or Credit & cash offices channel
type.
More than 50% refused loans belong to Credit & cash offices channel type.
Credit & cash offices channel type loans are getting canceled the most.
More than 90% Unused Offer loans belong to Country-wide channel type.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 119/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [116]: plot_pie_merge(appdata_merge,'NAME_YIELD_GROUP')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 120/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 121/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS NAME_YI

Refused A

28.5%
32.2%
34.7%

26.9% 9.98% 26

Canceled Un

21%
31.5%

49% 7.93%

9.84%

9.66%

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 122/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Most of the approved loans have medium grouped interest rate.

Loans with low or normal interest rate are getting refused or canceled the most.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 123/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [117]: plot_pie_merge(appdata_merge,'NAME_HOUSING_TYPE')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 124/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 125/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS NAME_HO

1.43% Refused 1.42% Appro

0.817% 0.826%

4.35%
4.63%
0.269% 0.315%

3.4

3.7
8%

4%
89.4%

1.3% Canceled Unused

0.742% 1.37%
3.94%

0.242% 1.22%
3.5

6.31%

3.0
0.782%
3%

8%

90.2%

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 126/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 127/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [118]: plot_merge(appdata_merge,'Age Group')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 128/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 129/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS Age

Refused

26.4% 26.7%
60k 25.7
24.7%
200k
22.6%
50k

40k 150k

14.5%
30k
11.8% 100k

20k

50k
10k

0 0
30-40 40-50 50-60 20-30 60-100 0-20 30-40 40-5

Canceled
8000
25.2% 24.8% 34.3%
60k 24.0%

50k 6000
26.3

40k
15.0%
4000
30k
11.0%
20k
2000

10k

0 0
40-50 50-60 30-40 60-100 20-30 0-20 30-40 40-5

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 130/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 131/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [119]: plot_merge(appdata_merge,'Work Experience')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 132/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 133/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS Work E

Refused
350k

54.4%

50.1%
100k
300k

80k
250k

60k 200k

28.3%
26.8%
150k
40k
100k

20k
10.4%

50k
4.0%
2.1% 1.1% 1.1%
0 0
0-5 5-10 10-15 15-20 20-25 25-30 30-100 0-5 5-10

Canceled
52.9%

100k
10k 53.6%

80k
8k

60k
6k
28.1%
26.9%

40k 4k

20k 2k
11.2%

4.3%
2.2% 1.3% 1.2%
0 0
0-5 5-10 10-15 15-20 20-25 25-30 30-100 0-5 5-10

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 134/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 135/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [120]: plot_merge(appdata_merge,'AMT_CREDIT_Range')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 136/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 137/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS AMT_CR

Refused

22.9% 21.2% 2
50k 21.1%
20.3% 20.1%
150k
40k
15.7%

30k 100k

20k
50k
10k

0 0
Medium Very Low Low High Very High Very Low

Canceled
5000
21.3% 21.8%
50k 20.6% 20.6% 2
20.1%

4000
17.5%
40k

3000
30k

2000
20k

10k 1000

0 0
Medium Very Low High Low Very High Very Low

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 138/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Most of the approved loans belong to Very Low & High Credit range.
Medium & Very Low credit range loans are canceled and rejected the most.

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 139/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

In [121]: plot_merge(appdata_merge,'AMT_INCOME_TOTAL_Range')

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 140/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 141/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

NAME_CONTRACT_STATUS VS AMT_INCOM

Refused
70k 250k
27.9% 27.6%
60k 26.1% 2
200k
50k

18.3% 150k
40k
15.6%
30k
12.1% 100k

20k
50k
10k

0 0
High Low Very High Very Low Medium Low

Canceled

70k 28.0% 28.1%

6000 2
26.6%
60k
5000

50k
4000
17.9%
40k
15.1%
3000
30k 12.4%
2000
20k

10k 1000

0 0
High Low Very High Very Low Medium Low

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 142/143

4/9/23, 6:41 PM EDA - Jupyter Notebook

Insights

Most of the loans are getting approved for Applicants with Low Income range. May be they are
opting for low credit loans.
Almost 28% loan applications are either getting rejected or canceled even though applicant
belong to HIGH Income range. May be they have requested for quite HIGH credit range.

END

localhost:8888/notebooks/Documents/Python Notebooks/EDA.ipynb 143/143