1. The AI & Analytics Engine Knowledge Hub
  2. Build from Scratch: Classification and Regression guide
  3. Predictive Modeling: Train your own machine-learning model & make predictions

Deploy trained machine learning model

Deploy your trained machine learning model to make online predictions via API calls or use the what-if tool

In this article, you will deploy your trained machine learning model. Your deployed model will be hosted on PI.EXCHANGE's cloud, and will be ready to serve 24/7 to receive requests and generate predictions.

1. Select Deployments

Select DEPLOYMENTS in the left navigation, and click on DEPLOY MODEL.

Step 1: Select a Model

  • Model: Select the XGBoost Classifier model that had the better performance.
  • Trained on dataset version: Ignore as this relates to Continuous Learning.

Deployment model

Step 2: Select a deployment action

There are three options to deploy a model:

  • Deploy to a new endpoint

  • Reactivate an existing deployment - Only available if the current model already has deployments

  • Replace an existing endpoint - Only available if there are active deployments in the organization

When replacing an existing endpoint, the replaced deployment will be deleted, and the new model deployment will be associated with that existing endpoint. You can create another deployment on the same model to replace the deleted deployment.

Note: The deployment and endpoint name are automatically pre-filled with default values. You can also modify them.

Deploy Model
In case the actively deployed model quota is met, you can select an active deployment to deactivate to make room for the new deployment.

Note: You can only select active deployment that they have access to. If you can’t see a deployment of interest, contact your organization owner to deactivate it for you.

Model Deployment

Step 3: Click Deploy

  • Automatically deploy updated versions of the same model: Leave it switched off as it relates to Continuous Learning.

  • Click Deploy to deploy the model

Note: An endpoint is like a URL assigned to your deployment, which is where your prediction request should send to.

Congratulations, you have successfully trained and deployed your very own machine learning model!