Credit Card Use: Consider the following hypothetical bank data on consumers use of credit card credit facilities in Table 11.3. Create a JMP data table using Table 11.3 (File>New>Data Table), and create a neural network model (Analyze>Predictive Modelling>Neural) using your new data table. Use the default validation method (Holdback Portion) and use the random seed 123. (see the screen capture below)

• What is the misclassification rate on the validation set? [ Select ] ["0", "0.4268", "0.012", "0.4"]
• How many parameters are estimated to build this model? [ Select ] ["14", "15", "13", "12"] (Hint: use the red triangle option "Show Estimates")

Use the Profiler (red triangle option) to explore how the predicted response changes as you change values of the predictors. Describe what you observe. Determine whether the statement below is either true or false.

• As years increases, the probability of "Use Credit = 1" decreases. As Salary increases, the probability of "Use Credit = 1" increases.. The highest probability that "Use Credit = 1" is when Years is low and Salary is High. (TRUE/FALSE?):

Fit another neural network to this same data. To produce the same results first set the random seed to 456 (in the first launch window). There will be differences between this model and the first model. Why do the results differ? Determine whether the statement below is either true or false.

• The starting values for the weights are different (random starting values are used unless the random seed is set), leading to different final models. (TRUE/FALSE?):

TABLE 11.3 Data for Credit Card Example and Variable Descriptions Years Salary Used Credit 43 65 4 18 1 3 15 6 53 95 88 112 0 1 0 0 1 Note : Years = Number of years that a customer has been with the bank; Salary = customer's salary (in thousands of dollars); Used Credit 1 = customer has left an unpaid credit card balance at the end of at least one month in the prior year; and 0 = balance was paid off at the end of each month. Fleidt toes Rows Cold Dot Analyse Graph - Untitled 10 Year Salary e Credit Col DOE Analyze Graph Tools Window He + 2 Select Columns 3 Columns www.ution of the Cast Selected Columns Roles Cano 4 395 5 1581 6 6 1121 move Informatie ing Set Random Seed 120 Becal - Neural Model Launch Von Method Product Holdback Proportion 0.3333 Random Seed Hoden Layer Structure Number och inte Activation Sigmoid Identity Radial Tante Gaussian 123 Columns Stary Credit Second Bootie Redes by the langue Number of Models Learning Bote Fring Options Transform Coveries Paty Method Number of Tours Go TABLE 11.3 Data for Credit Card Example and Variable Descriptions Years Salary Used Credit 43 65 4 18 1 3 15 6 53 95 88 112 0 1 0 0 1 Note : Years = Number of years that a customer has been with the bank; Salary = customer's salary (in thousands of dollars); Used Credit 1 = customer has left an unpaid credit card balance at the end of at least one month in the prior year; and 0 = balance was paid off at the end of each month. Fleidt toes Rows Cold Dot Analyse Graph - Untitled 10 Year Salary e Credit Col DOE Analyze Graph Tools Window He + 2 Select Columns 3 Columns www.ution of the Cast Selected Columns Roles Cano 4 395 5 1581 6 6 1121 move Informatie ing Set Random Seed 120 Becal - Neural Model Launch Von Method Product Holdback Proportion 0.3333 Random Seed Hoden Layer Structure Number och inte Activation Sigmoid Identity Radial Tante Gaussian 123 Columns Stary Credit Second Bootie Redes by the langue Number of Models Learning Bote Fring Options Transform Coveries Paty Method Number of Tours Go