Making predictions programmatically

The possibilities that programmatically using DataRobot presents are enormous. By using its API, models can be deployed and predictions can be made against them. Before making programmatical predictions within the production environment, models need to be deployed. DataRobot models are deployed using Portable Prediction Servers. These are Docker containers that can host machine learning models, which serve predictions and prediction explanations through a REST API.

To deploy models, we can use the DataRobot package's deployment method. Here, we must provide a description, the DataRobot model's ID, as well as its label to create the deployments. A typical Python deployment script follows this format:

deployment = dr.Deployment.create_from_learning_model(
MODEL_ID, label='DEPLOYMENT_LABEL', description='DEPLOYMENT_DESCRIPTION',
    default_prediction_server_id=PREDICTION_SERVER_ID)
deployment...