a. a lending firm has been using a decision tree (DT) based classifier to help determine which applicants are credit worthy (class 1) and which ones are high risk (class 2). The DT classifier has been found to make correct predicitions 80% of the time. An analyst has developed a new random forest (RF) based classifier that is claimed to e 85% accurate for both classes of applicants. On the other hand, the RF-based classifier is accurate for creditworthy applicants 90% of the time, but accurate for high-risk applicants 70% of the times. the cost (average loss from opportunity cost) of misclassifying an applicant who is truly creditworthy as a high risk one is $100, and the cost (average loss from potential default) of misclassifying an applicant who is truly high risk as a creditworthy one is also $100. Correct classifications do not lead to any loss for the firm. 75% of the applicants are truly creditworthy, while the remaining 25% are high risk. Given these numbers, which classifier should the firm use from an overall expected cost standpoint? Provide all details.

b. because of potential liability concerns, the audit department wants to make sure that all loans approval decisions can be justified if needed. Which of the two approached (decision tree based or Random Forecast based) would be preferable based on the requirement? Why?