10.2 Configuration

Once model is trained and deployed into production, you need to perform the following configuration in the production sanctions OFSAA application to ML integration for the batches.
  1. Create a same user in the OFSAA app as in Compliance Studio with CS Admin Access group.

    Figure 10-1 User Administrator group


    This image displays the CS User Administrator groups

  2. Update MMG_SERVICE_URL in the nextgenemf_config table with the BASE_URL value of the nextgenemf_config table present in the compliance studio schema.
    1. On compliance studio schema, open the nextgenemf_config table and copy the BASE_URL.

      Figure 10-2 The Base URL on the Compliance studio schema


      The Base URL on the Compliance studio schema

    2. On Sanctions Schema, open the nextgenemf_config table and paste BASE_URL value to the MMG_SERVICE_URL value.

      Figure 10-3 The MMG Service URL on the Sanctions Schema


      The MMG Service URL on the Sanctions Schema

  3. Import the compliance studio_server certificate to sanctions java key store path.
    1. Navigate to <COMPLIANCE_STUDIO_INSTALLATION_PATH>/mmg-home/mmg-studio/ conf path, copy the studio_server.cer certification and paste it into sanctions OFSAA server.
    2. Run the below command to replace the placeholders:
      keytool -importcert -keystore <JAVA_HOME>/lib/security/cacerts -
      storepass changeit -alias studio_ml -file <studio_certificate_path>/
      studio_server.cer

      Note:

      Use the studio_ml as the alias name while creating the certificate.
  4. Following three ML Process are introduced in our batches:
    1. CS_ML_Batch_Process: This process internally calls the Compliance Studio production workspace, Aggregate events Task, and ML Scoring Task.

      For more information refer to the Batch Framework for Sanctions Event Scoring section in the OFS Compliance Studio Administration and Configuration Guide

      The OOB for this process, you need to configure manually after CS Event Creation process and before the CS ML Score Evaluation process.

    2. CS_ML_Score_Evaluation: This process calls the F_EVALUATE_CS_ML_EVENT_SCORE function to update the v_status_code column in the CS_ALERTS table to either ML Closed or ML Escalated based on the ML auto action threshold settings, and it will also add entries to the CS_ALERTS_ML_AUDIT table.

      The Out of Box (OOB) process, this preset in the CS DATA LOAD GENERATION BATCH before the CS_End_To_End_End_Batch process.

      Figure 10-4 ML process batches successful screen in Data Load Event Generation Batch


      ML process batches successful screen in Data Load Event Generation Batch

    3. Zipper ML Process: It contains two tasks, such as ZipperMLDataMovement & ZipperMLWorkflow.

      The ZipperMLDataMovement process calls the F_POPULATE_ZCS_ML_EVNT_DTLS function to add events score data into the fcc_cs_ml_evnt_score and fcc_cs_ml_evnt_score_dtls tables.

      The ZipperMLWorkflow process calls the F_CS_ML_WORKFLOW_TRIGGER function. This function changes the alerts status to either ML Closed or ML Escalated.

      The OOB for this process is provided in Oracle CS Zipper Processing batch and it is present just before the Assign alerts to queue tasks.

      Figure 10-5 Zipper ML process successful screen in Oracle_CS_Zipper_Processing Batch


      Zipper ML process successful screen in Oracle_CS_Zipper_Processing Batch

      Note:

      • If any one of the event is in Pending Review, the alert status is updated as New.
      • If all the events in alert are ML auto closed, the alert status is updated as ML Closed.
      • If any one of the event is ML escalated, and no other event is in Pending Review, the alert status is updated as ML Escalated
  5. In the Config Schema, Configure the aai_wf_global_settings as suggested below:
    CONCURRENT_INSTANCE_WAIT=200
    REQUEST_BATCH_SIZE=40
    REQUEST_THREAD_WAIT_TIME=2
    LOGGING_REQUIRED=N

    Figure 10-6 The aai_wf_global_settings screen


    The aai_wf_global_settings screen

  6. Once all the configuration steps are done, you can run the Data load event generation and Oracle CS Zipper processing batch.

Note:

Model training input data is available in the ML4AML_SES_EVENT_INPUT table. For more information, see the Sanctions Event Scoring section in the OFS Compliance Studio Administration and Configuration Guide and OFS Compliance Studio Use Case Guides.