Training the machine learning services model

You can train the machine learning services model, by using the training reports in the Admin UI.

Before you begin

Fix Pack 10 and later, Fix Pack 7 Interim Fix 4Ensure you have completed the Using fastText word vectors task.

About this task

In the IBM® Product Master, there are three types of reports that can be used for training the ML services.
  • Machine Learning Services Training Report - Used for the initial training of the ML services. After successfully completion of this job, the data is inserted into the MongoDB collection and starts the ML services to start training.
    Table 1. Input parameter
    Container Name The name of the catalog.
    File location The location of the file that is used for training.
    • If the file location is System Directory, add the file name only.

      The file should be present in the following location.

      $TOP/supplier_base_dir/<company_name>/ctg_files/<file_name>
      Example

      catalog_data.xlsx

    • If the file location is Docstore, add the relative path of the file.
      Example

      /archives/catalog_data.xlsx
    Training samples file The file used for training.
    Service type The type of training (categorization).
    Table 2. Output
    report.out file A report.out file that contains a message.
    Example
    {"detail": "Started <ServiceType> training for model: <CompanyCode>_<CatalogName>_<ServiceType>."}
  • Machine Learning Active Services Report - Used to monitor training jobs and fetch list of all models that are successfully deployed by ML services along with the version, URL, status, and port number details. Models whose status is deployed are successfully trained. Other status can be training_inprogress and training_initiated.
    Table 3. Output
    report.out file A report.out file that contains a message.
    Example
    { "services": [ { "name": "<CompanyCode>_<CatalogName>_<ServiceType>", "version": <Version>, "status": "<Status>", "category": "<ServiceType>" } ] }
  • Machine Learning Services Retraining Report - Used for categorization retraining, the report fetches list of items that were mapped after the last execution of retraining job. If no retraining job is found, the report fetches the last initial training job. This report combines the items whose mapping has been updated in the collaboration area and items which you manually categorize due to incorrect prediction from ML service. After successful completion of this job, the data is added into the MongoDB collection with existing training data and ML training starts on entire data.
    Table 4. Input parameter
    Container Name The name of the catalog.
    Collaboration Name The name of the collaboration area.
    Training samples file The file used for training.
    Service type The type of training (categorization, attributes, and standardization).
    Table 5. Output
    report.out file A report.out file that contains a message.
    Example
    {"detail": "Started <ServiceType> training for model: <CompanyCode>_<CatalogName>_<ServiceType>."}

Procedure

To run a training report, proceed as follows:

  1. Log in to the Admin UI.
  2. Go to Product Manager > Reports > Reports Console.
  3. From Action column, click Report Console icon to run the report.
  4. Check the job status through the Schedule column. You can also check the status through Data model manager > Scheduler > Scheduler status.