A bank wants to optimize its loan underwriting procedures. Currently, it applies filters on loan applications that automatically reject the riskier ones. However, the bank is still approving too many applications that run into repayment issues.

The bank collects a large amount of data for each approved loan: $146$ fields. These fields can be split into a few distinct groups:

Loan demographics, such as the amount, the term, the interest rate, the reason for loan, and whether the person defaulted of their loan repayment.

Applicant demographics, such as age, salary, employment length, and home ownership.

Numerous risk factors, such as the number of public records, credit card delinquency, and bankruptcy. Roughly $70$ sparsely populated risk fields cover these risk factors.

Select the best problem statement

I want to build a model to determine if a loan will be a bad loan or a good loan, within $5$ days of receiving a loan application.

I want to determine the attributes about a person and loan type that are associated with a high probability of default $-$ and use them to build a model to predict whether a loan should be approved or denied by the loan underwriting team $-$ within $5$ days of receiving the loan application.

I want to determine the attributes about a person and loan type that are associated with a high probability of default $-$ and use them to build a model to predict whether a loan should be approved or denied by the loan underwriting team

I want to know what will cause a person to default on a loan