ATM Transaction of XYZ Bank¶

Exploratory Data Analysis¶

Amount is assumed to be in Dollars ($)

In [1]:

import pandas as pd
import plotly.express as px
import plotly.graph_objs as go
import plotly.offline as po
po.init_notebook_mode(connected=True)
import matplotlib.pyplot as plt
from matplotlib import rcParams
import seaborn as sb
%matplotlib inline
rcParams['figure.figsize'] = 8,5
# plt.style.use('fivethirtyeight')

In [2]:

atm = pd.read_csv('AggregatedData.csv')
atm.tail()

Out[2]:

	ATM Name	Transaction Date	No Of Withdrawals	No Of XYZ Card Withdrawals	No Of Other Card Withdrawals	Total amount Withdrawn	Amount withdrawn XYZ Card	Amount withdrawn Other Card	Weekday	Festival Religion	Working Day	Holiday Sequence
11584	Big Street ATM	29-09-2017	137	34	103	468800	146200	322600	FRIDAY	H	H	WHH
11585	Mount Road ATM	29-09-2017	79	27	52	305100	172500	132600	FRIDAY	H	H	WHH
11586	Airport ATM	29-09-2017	117	77	40	709900	576800	133100	FRIDAY	H	H	WHH
11587	KK Nagar ATM	29-09-2017	76	48	28	408700	279900	128800	FRIDAY	H	H	WHH
11588	Christ College ATM	29-09-2017	143	61	82	700400	364200	336200	FRIDAY	H	H	WHH

In [3]:

atm.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 11589 entries, 0 to 11588
Data columns (total 12 columns):
 #   Column                        Non-Null Count  Dtype 
---  ------                        --------------  ----- 
 0   ATM Name                      11589 non-null  object
 1   Transaction Date              11589 non-null  object
 2   No Of Withdrawals             11589 non-null  int64 
 3   No Of XYZ Card Withdrawals    11589 non-null  int64 
 4   No Of Other Card Withdrawals  11589 non-null  int64 
 5   Total amount Withdrawn        11589 non-null  int64 
 6   Amount withdrawn XYZ Card     11589 non-null  int64 
 7   Amount withdrawn Other Card   11589 non-null  int64 
 8   Weekday                       11589 non-null  object
 9   Festival Religion             11589 non-null  object
 10  Working Day                   11589 non-null  object
 11  Holiday Sequence              11589 non-null  object
dtypes: int64(6), object(6)
memory usage: 1.1+ MB

Data Cleaning

In [4]:

# renaming columns for easy accessing
atm = atm.rename(columns={
    'ATM Name':'atm_name',
    'Transaction Date':'transaction_date',
    'No Of Withdrawals':'no_of_withdrawals',
    'No Of XYZ Card Withdrawals':'xyz_card_withdrawals',
    'No Of Other Card Withdrawals':'other_card_withdrawals',
    'Total amount Withdrawn':'total_amt_withdrawn($)',
    'Amount withdrawn XYZ Card':'amt_withdrawn_xyz($)',
    'Amount withdrawn Other Card':'amt_withdrawn_other($)',
    'Weekday':'wkday',
    'Festival Religion':'festival_region',
    'Working Day':'working_day',
    'Holiday Sequence':'holiday_sequence'
})

In [5]:

atm['transaction_date'] = atm['transaction_date'].apply(lambda x: x.replace('-', '/'))
# Change transaction date to date data type
atm.transaction_date = pd.to_datetime(atm.transaction_date, format='%d/%m/%Y')
# Change all weekdays to title case
atm['wkday'] = atm['wkday'].str.title()

Number of Withdrawals¶

There is a higher frequency for withdrawals between 100 - 130.
There is less than 100 count for withdrawals starting from 200 to 2500

In [7]:

import functions as fn
fn.hist(atm, 'no_of_withdrawals', 'Distribution of Number of Withdrawals')

Number of withdrawals for xyz cards and other cards¶

There is a higher frequency for other card withdrawals compared to xyz card withdrawals
The distribution is skewed to the right for both cards
There is a low frequency for number of withdrawals starting from 100

In [8]:

fn.hist(atm, ['xyz_card_withdrawals', 'other_card_withdrawals'], 'Distribution of Number of Withdrawals by Card Type')

Amount Withdrawn by Card type¶

There is a high frequency of money withdrawn with other card compared to xyz card
For both card, there is high frequency for amount between $0 - $0.5M

In [9]:

fn.hist(atm, ['amt_withdrawn_xyz($)', 'amt_withdrawn_other($)'], 'Amount Withdrawn by Card type')

Number of Withdrawal by ATM Name¶

The highest withdrawal was made from KK Nagar ATM, this contributes to 28.1% of the total witdrawal made across all ATM
Mount Road ATM is the second ATM with the highest number of withdrawal, 22.7%.
The least withdrawal was from Airport ATM, 14.3%

In [10]:

fig = px.histogram(atm, 'atm_name', 'no_of_withdrawals', histnorm='percent',
             text_auto='.1f', width=800, title='Number of withdrawals in percent for each ATM')
fig.update_yaxes(showticklabels=False)

Average Amount Withdrawn from each ATM¶

KK Nagar ATM has the highest average amount withdrawn, $782.1K, followed by Christ College ATM with $556.8K.
The least average withdrawal is from Big Street ATM, $308.6K

In [12]:

fig = px.histogram(atm, 'atm_name', 'total_amt_withdrawn($)', histfunc='avg',
             text_auto='.4s', width=800, title='Average amount withdrawn in from each ATM')
fig.update_yaxes(showticklabels=False)

Sweetviv¶

Sweetviz is used below for a brief EDA on the dataset.

Withdrawals are most made on weekends.

In [13]:

import sweetviz as sv
report = sv.analyze(atm)

                                             |          | [  0%]   00:00 -> (? left)

In [12]:

report.show_html('SweetvizEDA.html', open_browser=False)

Report SweetvizEDA.html was generated.

In [13]:

report.show_notebook(scale=0.6, layout='widescreen', h='1400', w=1200)

Card Withdrawals by Weekday¶

There are more withdrawals on weekends than weekdays. The highest withdrawals are on Sundays.
Across all weekday, there is more withdrawals from xyz card than other card.

In [14]:

days = ['Sunday', 'Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday']
wk_grp = atm.groupby('wkday').sum().reindex(days)
wk_grp.reset_index(inplace=True)
fig = px.bar(wk_grp, 'wkday', ['xyz_card_withdrawals', 'other_card_withdrawals'], 
       width=800, barmode='group', text_auto='.3s', title='Card withdrawals by Weekday for each Card type')
fig.update_traces(textposition='outside')
fig.update_yaxes(showticklabels=False)

Amount Withdrawn by Weekday¶

Amount withdrawn from other card is low compared to amount withdrawn from XYZ card.
For both cards, the least amount withdrwan is on Friday.

In [41]:

fig = px.bar(wk_grp, 'wkday', ['amt_withdrawn_xyz', 'amt_withdrawn_other'], 
       width=800, barmode='group', text_auto='.3s', title='Amount withdrawn by Weekday for each Card type')
fig.update_traces(textposition='outside')
fig.update_yaxes(showticklabels=False)

Card Withdrawals from each ATM¶

In [15]:

atm_grp = atm.groupby('atm_name').sum()
atm_grp.reset_index(inplace=True)
fig = px.bar(atm_grp, 'atm_name', ['xyz_card_withdrawals', 'other_card_withdrawals'], barmode='group',
       width=900, text_auto='.3s', title='Card withdrawals from each Atm')
fig.update_traces(textposition='outside')
fig.update_yaxes(showticklabels=False)

In [19]:

from pandas_profiling import ProfileReport

In [20]:

profile = ProfileReport(atm, title='Pandas Profiling')

In [21]:

profile.to_widgets()

Summarize dataset:   0%|          | 0/5 [00:00<?, ?it/s]

Generate report structure:   0%|          | 0/1 [00:00<?, ?it/s]

Render widgets:   0%|          | 0/1 [00:00<?, ?it/s]

VBox(children=(Tab(children=(Tab(children=(GridBox(children=(VBox(children=(GridspecLayout(children=(HTML(valu…

In [22]:

profile.to_notebook_iframe()

Render HTML:   0%|          | 0/1 [00:00<?, ?it/s]

In [57]:

px.line(atm, 'transaction_date', 'no_of_withdrawals', color='atm_name', width=1000)

In [58]:

px.line(atm, 'transaction_date', 'no_of_withdrawals', facet_col='atm_name', facet_col_wrap=2,
        height=800, width=700)

In [96]:

new_df = atm.set_index('transaction_date')
new_df = new_df.resample('D').sum()
px.line(x=new_df.index, y= new_df['no_of_withdrawals'])

In [79]:

plt.style.use('seaborn')
# plt.style.available

In [80]:

plt.figure(figsize=(10,6))
plt.plot(new_df.index, new_df.xyz_card_withdrawals, '--', c='b')
plt.plot(new_df.index, new_df.other_card_withdrawals, '--', c='r')
# plt.grid()
plt.xlabel('Year')
plt.ylabel('No. of Withdrawals')
plt.legend(['XYZ Card', 'Other Card'])
plt.show()

In [90]:

plt.figure(figsize=(10,6))
plt.plot(new_df.index, new_df.amt_withdrawn_xyz, '--', c='b')
plt.plot(new_df.index, new_df.amt_withdrawn_other, '--', c='r')
# plt.grid()
plt.xlabel('Year')
plt.ylabel('Amount Withdrawan')
plt.legend(['XYZ Card', 'Other Card'])
plt.show()

In [86]:

monthly = new_df.resample('M').sum()
plt.figure(figsize=(10,6))
plt.plot(monthly.index, monthly.xyz_card_withdrawals, c='b')
plt.plot(monthly.index, monthly.other_card_withdrawals, c='r')
# plt.grid()
plt.xlabel('Year')
plt.ylabel('No. of Withdrawals')
plt.legend(['XYZ Card', 'Other Card'])
plt.show()

Out[86]:

	no_of_withdrawals	xyz_card_withdrawals	other_card_withdrawals	total_amt_withdrawn	amt_withdrawn_xyz	amt_withdrawn_other
transaction_date
2011-01-31	18012	10924	7088	64819400	44459200	20360200
2011-02-28	15637	9968	5669	57395900	40790200	16605700
2011-03-31	17727	11797	5930	66051300	48823000	17228300
2011-04-30	15784	9778	6006	59207400	40749000	18458400
2011-05-31	17154	10306	6848	65967100	43767100	22200000

In [91]:

plt.figure(figsize=(10,6))
plt.plot(monthly.index, monthly.amt_withdrawn_xyz, c='b')
plt.plot(monthly.index, monthly.amt_withdrawn_other, c='r')
# plt.grid()
plt.xlabel('Year')
plt.ylabel('Amount Withdrawan')
plt.legend(['XYZ Card', 'Other Card'])
plt.show()

In [109]:

atm1 = atm.sample(frac=0.1).copy()
plt.scatter(x=atm1['xyz_card_withdrawals'], y=atm1['other_card_withdrawals'])

Out[109]:

<matplotlib.collections.PathCollection at 0x2a708696a30>

In [102]:

g = sb.PairGrid(data=atm, vars=['no_of_withdrawals', 'xyz_card_withdrawals', 'other_card_withdrawals', 
                                'total_amt_withdrawn', 'amt_withdrawn_xyz', 'amt_withdrawn_other'],
                hue='atm_name')
g = g.map_offdiag(sb.scatterplot)
g.map_diag(plt.hist)
g.add_legend();

In [105]:

g = sb.PairGrid(data=atm, vars=['no_of_withdrawals', 'xyz_card_withdrawals', 'other_card_withdrawals', 
                                'total_amt_withdrawn', 'amt_withdrawn_xyz', 'amt_withdrawn_other'])
# g = g.map_offdiag(sb.scatterplot)
g.map_upper(sb.scatterplot)
g.map_lower(sb.kdeplot)
g.map_diag(sb.kdeplot)
g.add_legend();