Lending club case study
Group Members:
Mohit Kumar Dubey
Raghuveer Kona
Agenda
- This company is the largest online loan marketplace, facilitating personal
loans, business loans, and financing of medical procedures.
- Like most other lending companies,not lending loans and lending loans to
‘risky’ applicants are the largest source of financial loss (called credit loss).
- The company wants to understand the driving factors (or driver variables)
behind loan default, i.e. the variables which are strong indicators of default.
- The company can utilise this knowledge for its portfolio and risk assessment
Problem solving approach
Data Cleaning Data Derive columns Univariate Bivariate Observations
Standardisation analysis analysis
Observe and
Remove the more - Standardise the data - Derive new Analyse each Analyse the understand all the
than 40% null by removing the columns to create column by plotting behaviour and relations between
values. Remove the extras from value and buckets with huge the distribution of relation between two columns and give
columns that will not standardising the data variation values of that columns and the recommendations to
contribute in the type - Derive new column and create correlation between reduce the loss
analysis - Remove the outliers columns from them
existing values (like
get month from date)
Data Cleaning Data Standardisation Derive Columns
Dropped columns based on below - Few columns were having extra - Created bins for few columns that
delimiter, on removing those, these has very high variation in value like
- Initial data was consisting of 111 columns became useful for annual_inc, loan_amnt, int_rate etc
columns analysis(such as percentage). - Created new columns with month and
- Around 57 columns were have more - Removed the outliers from year from issue_d column
than 40% of null values annual_inc.
- 9 columns were having only 1 value
- Around 15 columns were not
contributing much to the loan analysis
as the data had several data values
that could not be categorized,
consisted month or dates, were
identifiers for individual entries and
more.,
Observation on loan
- Number of loans each year has increased from ‘07 to ‘11
- Highest defaulters % w.r.t all loans granted can be observed for loans issued in ‘07, ‘11 and ‘08
- Though we cannot draw any conclusion from the data alone, Considering the market knowledge we
estimated that the reason could be the real state crisis that happened in year 2008
Observation on interest rates
- The count of loans granted is high for interest bucket 9%-17% and so is the count of defaulters as well.
- If we observe the highest count of defaulters is for 13%-17% bucket
- But if we compare the ratio of fully paid vs defaulters for various interest buckets. We can observe that the ratio of defaulter
% is increasing as the interest is increasing
- If the interest rate is high, the probability of a person defaulting is also high
Observation on loan term
- The count of short term loan (36 months) is high and so the defaulter if we compare
the count of defaulters to term then it is high for 36 months because approved loan
is high
- But if we compare the ratio of fully paid vs charged off then we can easily conclude
that the defaulter % is very high for 60 months loan term
- Users took high loan at high interest rate for longer term and then failed to pay back
Observation on home ownership
- The defaulter percentage is high for Rent and
Mortgage
- The data of Other is very small to draw any
conclusion
Observation on user grade
- As the grade decreases the ratio % of defaulters increases
- The count of defaulters is highest for B as the loans granted is also highest
for B
- Granting high loan to a low grade user is risky
Observation on public record bankruptcy
- As the bankruptcy record increases the chances of a loan being defaulted also increases (The data for bankruptcy 2 is very low
to conclude but the percentage distribution is very high)
- The count of defaulters is highest for 0 bankruptcy records because the loan granted is also highest to users with 0 bankruptcy
records
- More loan is requested from users with bankruptcy records and hence resulted in defaulters
Observation on purpose of loan
- Higher loan is provided to small businesses followed by debt
consolidation, credit card
- Loan amount for defaulters is high for small businesses followed by debt
consolidation, credit card
Observation on verification status
- More loans were given to not verified
- The count of defaulters is high for not verified, followed by Verified and then source verified
- The highest ratio defaulter percentage is for verified users
- More loans were provided to not verified users based on annual income
Observation on state loan
- CA, FL and NY has the highest defaulter count
Recommendation
Based on our observations made from the dataset we have observed that the following parameters might be one of the factors to be
considered for identifying loan defaults.
- Loans provided on high interest rate are more prone to defaulter.
- Loans provided on longer term (60 months) and high loan value with high interest rates are more prone to defaulter.
- Borrower with rent or mortgage home ownership are more prone to turning out to a defaulter. Lending club should avoid
giving high loans to this category.
- Grade could be a good source to validate users. Lending club should be more clinical when issuing loan to a lower grade as
lower the grade, higher the probability of turning out to be defaulter
- Public bankruptcy record is also a good metric to analysis. Users with bankruptcy records are more prone to defaulter
- Loans provided to small businesses, debt consolidation and credit card are more prone to defaulter. Lending club should
investigate more for loan request for these purposes. They should also issue either less or deny such loan requests
- More loans were provided to not verified users and hence it has more count. However more defaulter % (proportion) is when
the user’s verification status is verified.
- Loans provided to users from CA, NL, NY are more prone to defaulter
