You work for a bank. You have a labelled dataset that contains information on already granted loan application and whether these applications have been defaulted. You have been asked to train a…

Question

Options

• AIncrease the size of the dataset by collecting additional data.
• BTrain a linear regression to predict a credit default risk score.
• CRemove the bias from the data and collect applications that have been declined loans.
• DMatch loan applicants with their social profiles to enable feature engineering.

• A
  5% (1)
• B
  82% (18)
• C
  5% (1)
• D
  9% (2)

Explanation

Predicting credit default risk from a labeled historical dataset is a supervised learning problem, and linear regression is a straightforward, interpretable model that produces a continuous risk score - directly addressing the task of scoring new applicants. Option A is a distractor because you already have a labeled dataset; collecting more data may help eventually, but it is not the required next step when a usable dataset exists. Option C conflates survivorship bias (a real but secondary concern) with the immediate task - and "declined" applicants lack outcome labels (you don't know if they would have defaulted), making them unusable for supervised training without further assumptions. Option D is ethically and legally problematic in most jurisdictions (privacy regulations prohibit using social profile data for credit decisions) and is not a valid feature engineering approach in this context.

Memory tip: Map the problem type to the tool - you have labels (defaulted/not), you need a score (continuous output), so regression fits. The three wrong answers all delay or derail building the model: one collects unneeded data, one chases unobservable labels, and one breaks privacy law.

No community discussion yet for this question.