Fit a tree model of answered using all the available predictors. This is the model from the previous project; you are welcome to cut and paste your code.

Create a confusion matrix for this model, using the predict$()$ function with the type $=$ "class" argument. $($The default probability threshold used by the function is $\mathrm{.}5.)$

$1.$ What is the number of true positives $($TP$)?$

$2.$ What is the number of false positives $($FP$)?$

$3.$ What is the number of true negatives $($TN$)?$

$4.$ What is the number of false negatives $($FN$)?$

Note that due to mismatched factor levels the location of the cells in the confusion matrix may have gotten switched around. The numbers are accurate, but you must take care to identify which cells contain TP$,$ FP$,$ TN and FN$.$

Using the confusion matrix in the previous question how much profit $($revenue $-$ costs$)$ could be expected with these costs$-$benefits?

Cost$-$Benefit Matrix:

observed yes

observed no

predicted yes

$$75$

$-$$$25$

predicted no

Hint: multiply the counts in the confusion matrix cells by the the cost$-$benefit matrix cells. Note: profit should not be negative! Make sure that you have correctly identified the true positives and the false positives in your confusion matrix.

The second data set for this assignment, adviseinvest$\_$customer$\_$data.csv$,$ contains information on prospective customers.$$ There is no target variable in this dataset because these customers have not yet been called.

Steps:

Use the supervised model you created for the first question to predict a probability of answering for each prospective customer.

Assign a predicted class label of "yes" to every customer whose model$-$estimated probability of answering that is $\mathrm{.}3$ or greater.$($This is the optimal class decision threshold.$).$

Filter the dataset to contain$$ only prospective customers with a predicted class label of "yes" $($those whose predicted probability of answering is greater than $\mathrm{.}3).$ This is your contact list.

How many customers are on the contact list?

Note that it is not strictly necessary to assign the class label.$$ You could just filter the dataset to include only customers whose predicted probability of answering is $\mathrm{.}3$ or greater.

Develop recommendations for the client, the Director of Sales at AdviseInvest. What, concretely, would you recommend that AdviseInvest do$?$ How should your analytical results be operationalized?

Write a paragraph or two in which you clearly describe and argue for your recommendations. You will be graded on the extent to which you are able to:

Describe the business problem and persuasively present your results as a solution, discussing the details of how your predictive modeling results would be used$/$implemented by the Director of Sales to improve operations.

Write clearly and compellingly without mistakes $($such as spelling errors$)$ and without relying on jargon. The Director of Sales is knowledgeable about analytics but is not an expert statistician. You should therefore clearly explain terms any technical terms you choose to use $($if any$).$

Feel free to produce additional analytics if you think it would help you make your case.