Training the Model Flow

When you train a model, you identify records as matches or non-matches by answering a series of questions.

After the completion of the Preparing Data stage, the model moves under the Training swim lane. At this stage, the model is ready for training. There can be variations where records are neither close to matches nor non-matches. Such records then become the input to the training process where you are prompted with questions seeking confirmation on whether a particular pair is a match or not. There is no specific sequence in which questions appear. You need to train the model flow to ensure that it can start matching. To train the model flow, perform the following steps:
  1. Under the Train swim lane, click the TRAIN button on the model flow that you want to train.

    The following image shows that the environment is setting up the training session:

    While the training session is being set up, you can navigate to other parts of the application by clicking on the link highlighted in the image above. If you navigate back to the application, the status of the model flow changes to Preparing training and then to Ready to train as shown in the following image:

    When the training session is ready, a notification appears with a link to start the training, as seen in the following image:

    Note: If you keep the training session idle for 30 minutes, then the environment is closed. You must restart the training.

    A series of questions appear and for each pair, you have to confirm whether the pair is the same or different. If you are not sure about the pair, you can also mark it as NOT SURE. The following images show some examples of questions:

    After answering some questions, if you continuously see questions showing pairs that are likely to be possible matches, you can end the training. At this point, the model flow is trained for matching.

  2. Click END TRAINING.

    The following image appears where you can choose to continue with the training or start matching:

  3. To end the training and start matching, click YES.
    The following image appears:

    At this point, the model is being prepared before it can start matching. Once the model is prepared, matching is triggered. Matching implies running the trained model against the tenant dataset to ensure that a user can review the match results to ascertain whether the model is able to deliver the expected results. This matching is different compared to matching by a published model. The match results by a published model appear on Hub associated with a Match IQ badge. When matching is complete, the model moves to the Review swim lane.

    Note: An email notification is sent to notify the designated user that the model is ready for review.
  4. If some questions are answered for a model but the training is not completed, then hover the mouse over the model under the Train swim lane, click the More options button, and select Finish Training.