Question 1. Suppose you are a sports agent negotiating a contract for Titus A., an athlete in the National Football League( NFL). An important aspect of any NFL contract is the amount of guaranteed money over the life of the contract. You have gathered data on 300 NFL athletes who have recently signed new contracts. Each observation (NFL athlete) includes values for:

  • percentage of his team’s plays that the athlete is on the field (SnapPercent),
  • the number of awards an athlete has received recognizing on-field performance (Awards),
  • the number of games the athlete has missed due to injury (GamesMissed),
  • millions of dollars of guaranteed money in the athlete’s most recent contract (Money, dependent variable).

You have trained the following decision tree to predict the dependent variable: (attaching image below) (referhwtress)

  1. Is this a regression or a classification tree?
  2. Titus’s variable values are: SnapPercent = 95, Awards = 6, and GamesMissed = 1. How much guaranteed money does the tree predict that a player with Titus’s profile should earn in his contract?
  3. Assume Titus feels that he was denied an additional award in the past season due to some questionable voting by some sports media. If Titus had won this additional award, how much money would the tree predict for Titus versus the prediction in part (b)? Comment on the result.
  4. According to the tree, what characteristics does an athlete must have to be offered the least amount of contract money.

  5. Question 2. Refer to the file named Cellphone.xlsx for this question.
  6. Make sure to read the contents of sheet “Description” for details of what the data is about.
    1. Import the data to JMP and set correct data types for each column.
    2. Create a decision tree classifier to predict the variable named Churn based on AccountWeeks, Contract Renewal, DataUsage, CustServCalls, DayMins, and Monthly Charge.
    • Create 3 splits.
    • Select Display Options > Show split prob; Show split count
    • Select Show Fit Details
    • Select Leaf Report
    • Don’t forget to save the script to the data table.
    • Inspect the output, and answer the following questions:
    1. List and interpret the set of rules given by the tree that characterizes churners.
    2. Is a customer who has an account for 40 weeks, renewed the contract recently, has a data usage of 5 Gb, made 6 calls to customer service, and the average number of daytime minutes per month is 350 predicted to churn or not? And with what probability?
    3. Compute and interpret the accuracy of the model.
    4. Compute and interpret the sensitivity.
    5. Compute and interpret the specificity.
    6. What percentage of predicted churners were false positives?
    7. What percentage of predicted non-churners were false negatives?