In this problem, you will develop a model to predict whether a person in the US Census earns more than $$50$K or not. Consider Income as the target variable and include Age, MaritalStatus, Race, Sex, and WeeklyHours as predictors. We use the Census dataset for this. Use a QDA model. Use the previously created $5$ folds for the cross$-$validation on the training set.

Calculate and show the confusion matrix for both the training and the test set. What is the performance with respect to qda$\_$model $<-$

# specify that the model is a quadratic discriminant analysis

discrim$\_$quad$()\%>\%$

# note: there are several potential engines for QDA, here we just use the default one

set$\_$engine$("$MASS$")\%>\%$

# select the binary classification mode

set$\_$mode$("$classification$")$

# then, let's put everything into a workflow

qda$\_$workflow $<-$ workflow$()\%>\%$

# add the recipe $($data pre$-$processing$)$

add$\_$recipe$($model$\_$recipe$)\%>\%$

# add the ML model

add$\_$model$($qda$\_$model$)$

set.seed$(1)$

control $<-$ control$\_$resamples$($save$\_$pred $=$ TRUE,

event$\_$level $=$ "second"$)$

qda$\_$fit $<-$

qda$\_$workflow $\%>\%$

fit$($data $=$ data$\_$train$)$

# investigate the result

qda$\_$fit

# to get the evaluation metrics for the test data:

qda$\_$final$\_$fit $<-$

qda$\_$workflow $\%>\%$

last$\_$fit$($data$\_$split$)$ # with the fit function, we train the model on the training data

# note that we use the test data here!

test$\_$predictions$\_$qda $<-$

qda$\_$final$\_$fit $\%>\%$

augment$()$

test$\_$predictions$\_$qda$Income $<-$ as$.$factor$($test$\_$predictions$\_$qda$Income$)$

# note: you need to select the truth and estimate variables based on the column names of the test object

classification$\_$metrics$($data $=$ test$\_$predictions$\_$qda,

truth $=$ Income,

estimate $=.$pred$\_$class,

$`.$pred$\_>50$K$`,$ # use the second outcome $($Yes$)$ as the level of interest

event$\_$level $=$ 'second'$)$ # note: the "second" indicates that we use the second class $($AHD $=$ Yes$)$ as the level of interest

# finally, let's create the confusion matrix and ROC curve

confusionMatrix$($data $=$ test$\_$predictions$\_$qda$$.$pred$\_$class,

reference $=$ test$\_$predictions$\_$qda$[[$target$\_$var$]],$

positive $=$ positive$\_$class$)$

two$\_$class$\_$curve$\_$test$\_$qda $<-$ roc$\_$curve$($data $=$ test$\_$predictions$\_$qda,

truth $=$ Income,

$`.$pred$\_>50$K$`,$

event$\_$level $=$ 'second'$)$

autoplot$($two$\_$class$\_$curve$\_$test$\_$qda$),$ sensitivity$,$ and specificity, and AUC? Create and print the ROC curves. I am using a five fold cross validation.

$******$How would I create the confusion matrix for the training set?$************$