Question 12
“Milo Loner (ML)” is a bank that gives out personal loans to customers. Currently, the process of document verification is done manually by verifying the scores from 2 variables “Earnings” and “Tenure”. The aim of inspection process is to identify customers who will default on a loan (considered as the negative class). From historical experience, manual inspection correctly identifies 85% of loan defaulters in any given batch of only defaulter customers.
The top brass at ML, Dr. Milo, has embarked on a journey in business analytics. Hence, he has decided to replace the manual inspection with an Automatic Defaulter Detection System (ADDS). This ADDS runs a trained logistic model in the background which classifies an applicant as “Defaulter” or “Non-Defaulter” based on the scores for “Earnings” and “Tenure”. To test the ADDS, a sample of 200 applications are taken. The sample is imbalanced and only 30% of the sample applicants were non-defaulters. The sample applications are passed through the ADDS, and the system identifies 70% of the actual non-defaulters as defaulters and 10% of actual defaulters as non-defaulters.
Using this information, answer the given subquestions.
Should ADDS be implemented?
Yes, the precision of ADDS in predicting defaulters is higher than the current manual inspection
No, the precision of ADDS in predicting defaulters is lower than the current manual inspection
Yes, the recall of ADDS in predicting defaulters is higher than the current manual inspection
No, the recall of ADDS in predicting defaulters is lower than the current manual inspection
Yes, the precision of ADDS in predicting non-defaulters is higher than the current manual inspection
No, the precision of ADDS in predicting non-defaulters is lower than the current manual inspection
Yes, the recall of ADDS in predicting non-defaulters is higher than the current manual inspection
No, the recall of ADDS in predicting non-defaulters is lower than the current manual inspection