The strategy table and strategy curve

Repeating the calculations you did in the previous exercise for several acceptance rates, you can obtain a strategy table. This table can be a useful tool for banks, as they can give them a better insight to define an acceptance strategy.

You know how to compute a bad rate for a certain acceptance rate by now, so the function strategy_bank was written and loaded into your workspace to speed things up. This function computes the cut-off and bad rate for the acceptance rates that are multiples of 5% (0%, 5%, 10%, …).

This exercise is part of the course

Credit Risk Modeling in R

View Course

Exercise instructions

Have a look at the function strategy_bank.
The vector predictions_cloglog contains predicted probabilities of default using the cloglog model you used in chapter 2, the vector predictions_loss_matrixcontains the predicted probabilities of default using the pruned tree including a loss matrix (previously constructed in chapter 3). Apply function strategy_bank to each of the prediction-vectors, assign the name strategy_cloglog and strategy_loss_matrix respectively.
The strategy tables can be obtained using the object names in combination with $table.
The strategy curves have been plotted for you. The strategy curve of the tree model shows pretty strange behavior. Because of the structure of classification trees, you might have a bigger chance for weird "jumps" here. Additionally, the tree with a loss matrix was a very big one, so this might be the result of overfitting!

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Have a look at the function strategy_bank


# Apply the function strategy_bank to both predictions_cloglog and predictions_loss_matrix



# Obtain the strategy tables for both prediction-vectors


# Plot the strategy functions
par(mfrow = c(1,2))
plot(strategy_cloglog$accept_rate, strategy_cloglog$bad_rate, 
     type = "l", xlab = "Acceptance rate", ylab = "Bad rate", 
     lwd = 2, main = "logistic regression")

plot(strategy_loss_matrix$accept_rate, strategy_loss_matrix$bad_rate, 
     type = "l", xlab = "Acceptance rate", 
     ylab = "Bad rate", lwd = 2, main = "tree")

Edit and Run Code