Creating a New Model in an Existing Workspace
Creating a new model in an existing workspace can be done in two different ways. The first is from the home dashboard and the second is from the workspace dashboard in which you wish to create the new model.
Creating a Model from the Dashboard
A new model can be added to any existing workspace by click "Create a New Workspace".
From there switching tabs from "New Workspace" to "Existing Workspace" will let you train a new model.
Once you have selected an existing workspace the rest of the process is identical to creating a new workspace. You must upload a CSV or connect a database, then select the variables you are interested in.
Deploy a New Application
A new model can be created inside of a specific workspace by going to the Applications tab, and clicking "Deploy a New App".
From here you can create several new types of applications, but for now, we are interested in creating a new model.
Then we can give the new model a name and description, clicking "Continue" will bring you to the variable selection stage.
In this situation, since we are not uploading a new dataset, we have to manually input the variables we are interested in. The variable names should be written into the Input and Output sections and separated with a comma.
Once you have selected the variables you are interested click finish and it will create the model. Now all that's left to do is train the model.
Training the Model
When you return to the workspace dashboard on the left side you will see a list of applications. If you click on "Predictive Models" you will see every predictive model in that workspace.
Clicking on the one you just created will bring you to a new page which includes all the details about the model such as which variables it is using, it's accuracy and loss, and the feature importance.
From this page, you can also set up a new training job on any dataset that exists in the workspace.
Now that you have trained your new model the last step is to make predictions!
You have two options for making predictions. The first is to change the current model endpoint to the new model. The other is to create a new endpoint with the new model. For more information on how to create a new endpoint click here.
Similarly to the predictive model, in order to change the endpoint, we must open the endpoint using the menu on the left of your screen.
Once you have selected the endpoint you will see a screen describing the model endpoint.
Scrolling down you will see a box labelled Endpoint IO. Clicking on the Predictive Model section will open a drop-down menu of all available models which you can then select between. Once you have switched to your new model press "Save" and you can begin making predictions with your new Model. You can come back at any time and change between models. You do not need to change Output Warehouse or Output Table in this situation.
Congratulations! You have now successfully deployed a new model.