In this scenario, we will use the pre-existing model to produce required dataset based on our dataset without human output.
Using the model, we obtain additional columns, namely
mlClass (the decision made by the model) and
mlConfidence (the probability that the transaction is a Risk) and combine it with the human results.
Please see: credit with human and ml for an example.
We will run the BYOM_flow notebook.
Similarly to the AutoAI flow, we will start with default performance model
then define different values in the performance model, calculate the ROI and finally save the rules as JSON.
Follow the Watson Machine Learning section to see how we can control the decisions made by the model.