Tutorial - Step 6

Data Mining

[Previous]  [Next]

We now wish to use our data mining techniques to discover any patterns that may be present in the data. In particular, we would like to identify what makes some customers respond to the newsletter while others do not; we will do this by targeting those who responded - the YES class.

Using the discovery node

• Create a discovery node and link to the data source node. The stream should now look like this:

   
   

  

• Edit the discovery node's options. This activates the discovery wizard.
  

• In Step 1 of the wizard, make sure that the Single Experiment option is selected, then click on the Next button.
  

• In Step 2 of the wizard, set the following options:

  • Set the target class to "YES" in order to build rules that describe the class of customers who responded to the newsletter.
    
  • Set the maximum fields in rule option to 3, so that only simple rules are generated.
    
  • Set the accuracy threshold slider to 40 percent.
    
  • Leave the remaining options with their default values. The dialog should look like this:
   
   

  

   
   

• Click on the Finish button to close the discovery wizard.
  

• Run the stream from the discovery node. The dialog below will appear during WITNESS Miner's search for rules:

   
   

  

   
   

• Wait until the discovery engine terminates (or click on the Stop button after about 3000 proposed moves). When the engine has stopped, WITNESS Miner copies the best rule to the rule scratch pad, which is similar to the standard Windows clipboard.
  

• To view the contents of the rule scratch pad quickly, click on the edit menu, then select the view rule scratch pad command. Click OK to close the scratchpad.

[Previous]   [Next]