Recommendations for training machine learning models
To train a machine learning (ML) model, review the following recommendations:
•You can train an ML model with a subset of production data. Consider training a match model in a non-production environment that does not contain the entire production data.
•A match model that you want to use for training must have the same population as the production data.
•During the training process, the training page displays the precision, recall, and accuracy values that indicate the learning of the ML model. Consider reviewing the table in between the batches during the training process.
•When the precision, recall, and accuracy values display a steady increase for an ML model during the training of several batches, it indicates that the ML model is completely trained and you can stop training the ML model.
•For an optimal ML model training, you might want to train the model for at least 20 batches.