For this part, you will use JMP Pro to build and test five classifier models – KNN,
Partition (decision tree), Boosted Tree, and Neural Network. Follow the instruction given below.
You will use german-bank.jmp dataset for all five models. For more details about how to use each
classifier model, refer to Predictive and Specialized Modeling.pdf documentation.
(1). Naïve Bayes
Start JMP Pro
Open german-bank.jmp
Select Analyze > Predictive Modeling > Naïve Bayes
Select checking through foreign and click X,Factor
Select class and click Y,Response
Select Validation (on the left) and click Validation (on the right)
Click OK.
You will see Naïve Bayes Report (or output) that is similar to Figure 9.1 in Predictive and
Specialized Modeling.pdf documentation.
Capture the report screen and paste it in your submission.
The dataset’s class attribute has two possible values – 1 and 2.
Using the confusion matrix of Validation in the report (There are two confusion matrices.
Make sure that you use the Validation confusion matrix), calculate the following measures
for both classes (similar to those in Weka’s output window):
TP Rate FP Rate Precision Recall F-Measure Class
1
2
(2). KNN
Start JMP Pro
Open german-bank.jmp
Select Analyze > Predictive Modeling > K Nearest Neighbors
Select checking through foreign and click X,Factor
Select class and click Y,Response
Select Validation (on the left) and click Validation (on the right)
Click OK.
You will see K Nearest Neighbors Report (or output) that includes, among others, Model
Selection, Misclassification Rate for both Training and Validation, and confusion matrices
for best K value.
Capture the report screen and paste it in your submission.
What is the best K value?
The dataset’s class attribute has two possible values – 1 and 2.
Using the confusion matrix of Validation in the report (There are two confusion matrices.
Make sure that you use the Validation confusion matrix), calculate the following measures
for both classes (similar to those in Weka’s output window):
TP Rate FP Rate Precision Recall F-Measure Class
1
2
(3). Partition Model (decision tree)
Start JMP Pro
Open german-bank.jmp
Select Analyze > Predictive Modeling > Partition
Select checking through foreign and click X,Factor
Select class and click Y,Response
Select Validation (on the left) and click Validation (on the right)
Click OK.
In the output (this output is called platform report window), click Go. You will see a
decision tree in the output.
Click a red triangle next to Partition for Class and select Show Fit Details. Confusion
matrices will be added to the output.
Capture the output screen, which includes a decision tree and confusion matrices, and paste
it in your submission.
The dataset’s class attribute has two possible values – 1 and 2.
Using the confusion matrix of Validation in the report (There are two confusion matrices.
Make sure that you use the Validation confusion matrix), calculate the following measures
for both classes (similar to those in Weka’s output window):
TP Rate FP Rate Precision Recall F-Measure Class
1
2
(4). Boosted Tree
Start JMP Pro
Open german-bank.jmp
Select Analyze > Predictive Modeling > Boosted Tree
Select checking through foreign and click X,Factor
Select class and click Y,Response
Select Validation (on the left) and click Validation (on the right)
Click OK.
Gradient-Boosted Tree Specification window appears.
In the Reproducibility panel, select Suppress Multithreading and enter 123 for Random
Seed.
Click OK.
Capture the output screen, and paste it in your submission.
The dataset’s class attribute has two possible values – 1 and 2.
Using the confusion matrix of Validation in the report (There are two confusion matrices.
Make sure that you use the Validation confusion matrix), calculate the following measures
for both classes (similar to those in Weka’s output window):
TP Rate FP Rate Precision Recall F-Measure Class
(5). Neural Network
Start JMP Pro
Open german-bank.jmp
Select Analyze > Predictive Modeling > Neural
Select checking through foreign and click X,Factor
Select class and click Y,Response
Select Validation (on the left) and click Validation (on the right)
Click OK.
In the next window (it is called Model Launch Control Panel), verify that 3 is in the box
corresponding to TanH and First.
Click GO.
In the next screen, click red triangle next to Model NTanH(3) and select Diagram. You will
see confusion matrices and a neural network diagram in the output window.
Capture the output screen, and paste it in your submission.
How many hidden layers does the model have?
The dataset’s class attribute has two possible values – 1 and 2.
Using the confusion matrix of Validation in the report (There are two confusion matrices.
Make sure that you use the Validation confusion matrix), calculate the following measures
for both classes (similar to those in Weka’s output window):
TP Rate FP Rate Precision Recall F-Measure Class