{c} Use misclassication costs to develop the tree models {rpart and CSJCII} are the trees here different than ones obtained earlier? Compare performance of these two new models with those obtained earlier (in part 3a, b above}. Let's examine your 'best' decision tree model obtained. What is the tree depth? And how many nodes does it have? What are the important variables for classifing \"Good\" vs 'Ead" credit? Identify two relatively pure leaf nodes. What are the 'probabilities for 'Good' and 'Bad' in these nodes? The predicted probabilities can be used to determine how the model may be implemented. We can sort the data om high to low on predicted probability of \"good\" credit risk. Then, going down the cases from high to low probabilities, one may be able to determine an appropriate cutoE probability values above this can be considered acceptable credit risk. The use of cost gures given above can help in this analysis. For this, rst sort the validation data on predicted probability. Then, for each validation case, calculate the actual cost-benefit of extending credit. Add a separate column for the cumulative net cost-benet. How far into the validation data would you go to get maximum net benet? In using this model to score- fumre credit applicants, what cutovalue for predicted probability would you recommend? Provide appropriate performance values to back up your recommendation