Assignment #7 - Decision Tree

Assignment #6: Decision Tree Submission Instructions  Submit the following two files through Canvas>Assignments>To-Dos: (1) The completed, working Python script that produced the analysis with the minimum split set to 55 in Part 1. (2) The completed answer sheet provided on the last two pages as a separate file (for Q1- Q10 in Part 1 and Q11-Q13 in Part 2).  If you do not follow the instructions, your assignment will be counted late. o Late Assignment policy: Same as before. Evaluation Your submission will be graded based on the correctness of the completed answer sheet, with other files as supporting documents. Part 1. Decision Tree Analysis in Python Before you start Make sure you are using the most recent Python/Jupyter Notebook version! For this assignment, you’ll be working with the BankLoan.csv file and the Decision_tree.ipynb script (which we used in ICA #12). The BankLoan.csv file has data about 600 customers who received personal loans from a bank. The president of the bank wants to predict how likely a future customer is to pay back their loan so she can make better loan approval decisions. The data file contains the following variables: Variable Name Variable Description ID Customer identification number age The age of the customer, in years sex The gender of the customer region The type of area where the customer lives (INNER_CITY, TOWN, SUBURBAN, RURAL) income Customer’s yearly income in dollars married Whether the customer is married (0 = no, 1 = yes) children How many children the customer has car Whether the customer has a car (0 = no, 1 = yes) save_act Whether the customer has ever had a savings account (0 = no, 1 = yes) current_act Whether the customer has an active account (0 = no, 1 = yes) mortgage Whether the customer has a mortgage (0 = no, 1 = yes) payback Whether the customer paid back their loan (0 = no, 1 = yes) NOTE: payback is the outcome variable we are interested in here. It describes a categorical event (0 = no, 1 = yes). Page 1

Guidelines: 1) You’ll need to modify the script with the following information to perform the analysis:  Set the input filename to the bank’s dataset (i.e., BankLoan.csv).  Set the training partition (using TRAINING_PART) to 0.61 (61%) of the data set.  Set the minimum split (using MINIMUMSPLIT) to 55.  Make sure the outcome column setting is correct for your data set (using OUTCOME_COL).  You will need to modify the model to reflect the data set. This requires editing the cell 2 of the Decision_tree.ipynb script. Make sure you choose the correct outcome variable and exclude the variables that are inappropriate for the analysis. (HINT: ID is irrelevant to the analysis.) 2) Once you finish modifying the script, you can run the script. 3) Based on your script output, answer Questions 1-6 in the answer sheet at the end of this document: (NOTE: When asked “how likely…” cite the percentage!) 4) Now change the minimum split from 55 to 45 and re-run the script. Using the new tree, answer Questions 7-10 in the answer sheet at the end of this document. NOTE: Scientific Notations in Python In the output, you may see numbers like -1.8e-04 or 31e+5 . The "e" is a symbol for base-10 scientific notation. The "e" stands for ×10 exponent . So -1.8e-04 means −1.8×10 −4 . In fixed-point notation that would be -0.00018. Similarly, 31e+5 means 31×10 5 . In fixed-point notation that would be 3,100,000. Page 2

Answer Sheet for Assignment: Decision Trees in Python Name __________________________________ Fill in the answer sheet below. Question Answer Part 1. Decision Tree in Python (Minimum Split = 55) 1 How often will this tree make a correct prediction (include decimals)? Provide your answer for both the training set and the validation set. 2 How likely is a customer to pay back their loan if they are married, have one child, and make less than $30,000 in income? (NOTE: When asked “how likely…” cite the percentage!) 3 How likely is a customer to pay back their loan if they are not married, have one child, and a mortgage regardless of age? 4 How likely is a customer to pay back their loan if they are not married, have 2 children, and make $50,000 in income? 5 Describe the profile of the least likely customer to successfully repay their loan. 6 Describe the profile of the most likely customer to successfully repay their loan. (Minimum Split = 45) 7 How often will this new tree make a correct prediction (include decimals)? Provide your answer for both the training set and the validation set. 8 Is this model better or worse than the Page 4

first model at predicting who will repay their loan? Explain how changing the complexity factor affected the tree using no more than two sentences . 9 How likely is a customer to pay back their loan if they are married, have two children and make $35,000 per year? 10 Does having a saving account increase or decrease the likelihood that a customer will pay back their loan? Part 2 Compute and Evaluate Decision Trees 11 What is the correct classification rate for Tree #1? 12 What is the correct classification rate for Tree #2? 13 Which decision tree (Tree #1 versus Tree #2) has higher classification accuracy? Page 5