Serve Models to MLeap / Docker Containers

Description:

User can create/train model using the 'Train Model' processor within a workflow so that they can use the model effectively.

Features:

• To train a model from workflow, user can create a new model, as well as update the previously created models by creating new versions.
• Within the model view, user can set any one of the created versions to an active version.
• To create new models, a unique model name (within the domain) is auto-generated, and is editable so that user does need to bother finding an useful name unless required. 
• Models created can be shared within different projects so everyone can work with the model directly.
• Update/edit of metadata information is possible where a user can also re-name the model (unique name).
• User is able to delete and get rid of old/outdated models.
• User can see and update the meta-information by invoking the API from outside.

How it works:

1. Within a workflow, use a train model processor to either create new model or create new version of the existing model.

2. Upload a data-set and use horizontal split processor to split the data as input to 'Train Model'

3. Within the train model processor, a default selection is to 'Create new model'. For creating a new version, a drop down list with all the existing models is displayed.

Let us continue with the default selection and create a new model with the below configuration settings for 'Train Model' processor.

• Toggle the 'Decision Tree Classification' option (Check for other options as well)
• Click on 'Add Group'

• Fill the respective columns as:

  Dependent	category
  Independent	rank, sales
  Forecast column name	forecast_name
  seed	1
  Probability base column name	probability_name

  Save & Execute the workflow.

4. After executing the workflow, go to models tab where a model is created with the specified name.

The image attached below shows a trained model with different versions. (Check the checkbox for 'Active Version' to set any version as active).

5. Within the model, click on 'Open API Tester' and input values to test API and check the output section for results.