Skip to content

Develop a ML model that can accurately predict credit risk for loan applicants based on historical data and financial metrics.

Notifications You must be signed in to change notification settings

shemanto27/Credit-Risk-Modeling-on-Bank-Data-using-Machine-Learning

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 

Repository files navigation

Credit Risk Modeling of Banking Data

graphic_01-2

We have customer's data,we need to predict whether or not give a loan

Banking Terminologies

  • Asset: profit making products for Bank, simply loan/credits that will make profit for the bank.Loan product-Housing loan, Vehicle loan,Group loan, Education loan,Credit Card

  • Liabilities: Products for which bank is loosing money or has to give money to a person,loss making products for banks- Current account,Savings account(CASA), Fixed deposit(FD), Recurring Deposit(RD)

  • NPA: Non Performing Asset, loan that is default, loan account when DPD > 90 Days

  • Disbursed Amount: loan amount given to a customer

  • OSP(Outstanding Principle): let's say you took loan of 1 lack with EMI of 10k monthly, after 4 months you have paid 40k and remaining Bakaya/Balance is 60k that you have to pay. This Bakaya/Balance is OSP. OSP should be zero at the end of the loan cycle.

  • EMI(Equated Monthly Installment)

  • DSP(Days Past Due): Days I am late to pay. DPD ideally should be zero. when DPD become greater than zero, you become defaulted(you missed your EMI)

  • PAR(Portfolio at Risk): OSP when DPD > 0, they remaining money of bank that is at risk

  • Deliquency: Doing default

  • CIBIL: Credit Information Bureau India Limited. It's a credit information company that collects and maintains records of credit card-related activities for individuals and businesses.Records loan and credit card repayments.Provides credit scores and reports to lenders.CIBIL credit score ranges from 300 to 900.More your credit score, better you are

  • TL: Trande Line or loan or Assets

  • Secured TL: loan that is secured by collateral

Credit Risk Types

  • DPD(zero): NDA(Non deliquent account),person who never become default = No default account

  • DPD(0 to 30): SMA1(Standard Monitoring Account)

  • DPD(31 to 60): SMA2(Standard Monitoring Account)

  • DPD(61 to 90): SMA3(Standard Monitoring Account)

  • DPD(90 to 180): NPA

  • DPD(>180): Writen-off(Loan which is not present), bank will remove that NPA from book because,NPA improve = Loan portfolio quality of the bank will be better= Market sentiment will be good = stock price will improve

NPA Types

  • GNPA: Gross NPA (3-5%) = OSP default,means that bank gave loan of 100 tk and 2-5 tk could not recover

  • NNPA: Net NPA (0.01-0.06 %), Provisioning Amount subtracted

Understanding Dataset

we have two dataset, one CIBIL dataset that shows data of credit card users.second dataset is about the user that took any kind of product(loan) from bank

CIBIL Dataset

  • Target Column: Approved_Flag(P1-P4) shows priority level of an user. P1= Really good

CIBIL Dataset Columns

Column Name Description
time_since_recent_payment Time since recent payment made
time_since_first_deliquency Time since first delinquency (missed payment)
time_since_recent_deliquency Time since recent delinquency
num_times_delinquent Number of times delinquent
max_delinquency_level Maximum delinquency level
max_recent_level_of_deliq Maximum recent level of delinquency
num_deliq_6mts Number of times delinquent in last 6 months
num_deliq_12mts Number of times delinquent in last 12 months
num_deliq_6_12mts Number of times delinquent between last 6 and 12 months
max_deliq_6mts Maximum delinquency level in last 6 months
max_deliq_12mts Maximum delinquency level in last 12 months
num_times_30p_dpd Number of times 30+ DPD (days past due)
num_times_60p_dpd Number of times 60+ DPD (days past due)
num_std Number of standard payments
num_std_6mts Number of standard payments in last 6 months
num_std_12mts Number of standard payments in last 12 months
num_sub Number of sub-standard payments (not making full payments)
num_sub_6mts Number of sub-standard payments in last 6 months
num_sub_12mts Number of sub-standard payments in last 12 months
num_dbt Number of doubtful payments
num_dbt_6mts Number of doubtful payments in last 6 months
num_dbt_12mts Number of doubtful payments in last 12 months
num_lss Number of loss accounts
num_lss_6mts Number of loss accounts in last 6 months
num_lss_12mts Number of loss accounts in last 12 months
recent_level_of_deliq Recent level of delinquency
tot_enq Total enquiries
CC_enq Credit card enquiries
CC_enq_L6m Credit card enquiries in last 6 months
CC_enq_L12m Credit card enquiries in last 12 months
PL_enq Personal loan enquiries
PL_enq_L6m Personal loan enquiries in last 6 months
PL_enq_L12m Personal loan enquiries in last 12 months
time_since_recent_enq Time since recent enquiry
enq_L12m Enquiries in last 12 months
enq_L6m Enquiries in last 6 months
enq_L3m Enquiries in last 3 months
MARITALSTATUS Marital status
EDUCATION Education level
AGE Age
GENDER Gender
NETMONTHLYINCOME Net monthly income
Time_With_Curr_Empr Time with current employer
pct_of_active_TLs_ever Percent active accounts ever
pct_opened_TLs_L6m_of_L12m Percent accounts opened in last 6 months to last 12 months
pct_currentBal_all_TL Percent current balance of all accounts
CC_utilization Credit card utilization
CC_Flag Credit card flag
PL_utilization Personal loan utilization
PL_Flag Personal loan flag
pct_PL_enq_L6m_of_L12m Percent enquiries PL in last 6 months to last 12 months
pct_CC_enq_L6m_of_L12m Percent enquiries CC in last 6 months to last 12 months
pct_PL_enq_L6m_of_ever Percent enquiries PL in last 6 months to ever
pct_CC_enq_L6m_of_ever Percent enquiries CC in last 6 months to ever
max_unsec_exposure_inPct Maximum unsecured exposure in percent
HL_Flag Housing loan flag
GL_Flag Gold loan flag
last_prod_enq2 Latest product enquired for
first_prod_enq2 First product enquired for
Credit_Score Applicant's credit score
Approved_Flag Priority levels

Bank's Internal Product Dataset

  • Target Column:

Bank's Internal Product Columns

Column Name Description
Total_TL Total trade lines/accounts in Bureau
Tot_Closed_TL Total closed trade lines/accounts
Tot_Active_TL Total active accounts
Total_TL_opened_L6M Total accounts opened in last 6 months
Tot_TL_closed_L6M Total accounts closed in last 6 months
pct_tl_open_L6M Percent accounts opened in last 6 months
pct_tl_closed_L6M Percent accounts closed in last 6 months
pct_active_tl Percent active accounts
pct_closed_tl Percent closed accounts
Total_TL_opened_L12M Total accounts opened in last 12 months
Tot_TL_closed_L12M Total accounts closed in last 12 months
pct_tl_open_L12M Percent accounts opened in last 12 months
pct_tl_closed_L12M Percent accounts closed in last 12 months
Tot_Missed_Pmnt Total missed payments
Auto_TL Count of automobile accounts
CC_TL Count of credit card accounts
Consumer_TL Count of consumer goods accounts
Gold_TL Count of gold loan accounts
Home_TL Count of housing loan accounts
PL_TL Count of personal loan accounts
Secured_TL Count of secured accounts
Unsecured_TL Count of unsecured accounts
Other_TL Count of other accounts
Age_Oldest_TL Age of oldest opened account
Age_Newest_TL Age of newest opened account

About

Develop a ML model that can accurately predict credit risk for loan applicants based on historical data and financial metrics.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published