Importing and running the model

You add the Machine Learning model Mortgage Approval Prediction Model to the Miniloan decision service in Rule Designer, and run the application.

To import the model and run it in the Miniloan sample:

1. Open the Miniloan sample in Rule Designer (see Task 1: Designing the main rule project in the decision service) to enhance it with Machine Learning.
2. Next, see Integrating Machine Learning for details on importing the model. Use the deployment ID you obtained in step 19 of Downloading the Machine Learning model.
3. Replace the contents of the ScoringML_1_Call action rule with the following code:

   Then 
   set the INCOME of 'the scoring' to 100000; 
   set the YRS AT CURRENT ADDRESS of 'the scoring' to 5; 
   set the YRS WITH CURRENT EMPLOYER of 'the scoring' to 320; 
   set the NUMBER OF CARDS of 'the scoring' to 1; 
   set the CREDITCARD DEBT of 'the scoring' to 1000; 
   set the LOAN AMOUNT of 'the scoring' to 30; 
   set the CREDIT SCORE of 'the scoring' to 300; 
   set the PROPERTY VALUE of 'the scoring' to 750000; 
   set the AREA AVG PRICE of 'the scoring' to 600000; 
   set the LOANS of 'the scoring' to 0; 
   set the GENDER of 'the scoring' to "M"; 
   set the EDUCATION of 'the scoring' to "1"; 
   set the EMPLOYMENT STATUS of 'the scoring' to "1"; 
   set the MARITAL STATUS of 'the scoring' to "1"; 
   set the APPLIEDONLINE of 'the scoring' to "1"; 
   set the RESIDENCE of 'the scoring' to "1"; 
   set the COMMERCIAL CLIENT of 'the scoring' to "0"; 
   set the COMM FRAUD INV of 'the scoring' to "0"; 
   set the deployment ID of 'the scoring' to 'the deploymentID in variable set scoringModel';
   make it false that 'the scoring' is permitted to use defaults ; 
   execute 'the scoring' ; 
   add "ML return code=" + the return code of 'the scoring' to the messages of 'the loan'; 
   add "Items of the list probabilities:" to the messages of 'the loan'; 
   for each number in the probabilities of 'the scoring' : - add "- " + this number to the messages of 'the loan'; 
   add "prediction=" + the prediction of 'the scoring' to the messages of 'the loan'; 

   The code changes the default values to valid values for a mortgage application, for example:

   • set the GENDER of 'the scoring' to "M";
   • set the AREA AVG PRICE of 'the scoring' to 600000;

   The last four lines of the code allow the Machine Learning decision to be passed back via the messages of the loan.

4. Replace the contents of the ScoringML_2_Decision action rule with the following code:

   if the prediction of 'the scoring' is at most 10 
   then add "prediction <= 10" to the messages of 'the loan' ; 
   else add "prediction > 10" to the messages of 'the loan' ; 

5. Integrate the Machine Learning rules into the miniloan ruleflow:
   1. Open the miniloan ruleflow in the ruleflow editor.
   2. Drag and drop the rules ScoringML_1_Call and ScoringML_2_Decision into the editor.
   3. Move the end of the data approved line to ScoringML_1_Call.
   4. Connect ScoringML_1_Call to ScoringML_2_Decision.
   5. Connect ScoringML_2_Decision to the eligibility task.

   The modified ruleflow should look as follows:

   
6. If you have not already done it, you can configure your Liberty server with the details of the Machine Learning endpoint (see Connecting to the Machine Learning service). Use the URL in step 18 of Downloading the Machine Learning model.
7. Deploy the rule project to the Rule Execution Server console in the usual manner (see Task 5: Deploying and running your ruleset).
8. In the decision service HTDS endpoint, change <approved>false<approved> to <approved>true<approved> in the default input to get the extra loan messages in the output.
9. Click Execute Request with the altered input (see step 5 of Task 5: Deploying and running your ruleset).
   The output includes Machine Learning messages: