Algorithms

Get steps on how to do Algorithm Settings and Project level access to use Algorithms inside Notebook.

Pre-requisite:

1. Configure the Algorithms using the Admin module to access them under the Data Science Lab Project creation.

2. The user must select Algorithms while creating a Project to make them accessible for a Notebook within the Project.

Check-out the walk through on how to apply Algorithms inside Notebook.

Algorithm tab in the Notebook

The entire process to access the Algorithms option inside the DS Lab and actually create a model based on the Algorithm is a three-step process:

1. Admin Settings for Algorithms

2. Project Level Algorithm Selection

3. Using Algorithm inside a Notebook

Admin Settings for Algorithms

• Navigate to the Admin module.

• Open the Notebook Settings option from the Configuration section of the Admin panel.

• The Notebook Settings page opens.

• Select the Algorithms using the drop-down option.

• Click the Save option.

• A confirmation message appears to inform about the Notebook details updates.

Project Level Algorithm Selection

Once the Algorithm settings is configured in the Admin module, the user can access the Algorithm within the Data Science Lab.

• Navigate to the Data Science Lab.

• Click the Create Project option.

• Together with the other required fields select the algorithms using the given checkboxes from the drop-down menu.

Selecting Algorithms using the Project page

• The selected Algorithms appear on the field separated by comma.

Selected Algorithms

• Save the project.

Saving the Project with Algorithms

Using Algorithms inside a Notebook

Once the Algorithms are selected while creating a Project, those algorithms will be available for all the Notebooks created inside that project.

• Open the Notebook tab inside the same Project.

• Create a new Notebook or Navigate to an existing Notebook under the same Project.

• You can see the selected Algorithms at the project level get listed under the right-side panel of the Notebook.

Accessing the selected Algorithms inside a Notebook

• Add a code cell with the dataset information.

• Define the Dataframe using another code cell.

Code cells with Dataset and Data frames

Please Note: You can run the cell containing the data frame details to see the data output below:

• Add another code cell below.

• Select an algorithm type by using the checkbox.

• You will get the algorithm script pasted in the intended code cell.

Default script gets added to the Code cell based on the selected Algorithm type

• Specify train and test data.

• Run the cell.

Specifying the train and test data

• You can see the data output below.

• Add a new code cell below.

• Click the Save Model option.

• Specify the model name and model type in the auto generated code.

• Run the cell.

Saving the Algorithm based model

• The model gets saved and lists under the Models tab (The model will list under the Unregistered models as it is not registered yet).

The saved model

• Add new code cell.

• Load the model in the cell by clicking the checkbox given next to the model.

The saved model gets loaded in the new cell

• Specify the model as the loaded model and model type for the loaded model.

• Run the cell.

Specifying the model and model type for a loaded model

• The data output gets displayed below.

Output data after running the loaded model

Registering an Algorithm based Model

This section displays the flow on how to register a saved model from a Notebook.

• Click the Register option for the saved model.

• The Register Model dialog box appears to confirm the action.

• Click the Yes option.

• A confirmation message appears to inform about the completion of the model registration action.

• The concerned model appears under the Registered model list.

• Open a Pipeline using the Data Pipeline module that contains the DS Lab Model Runner component.

• You can see the same model (which was registered from the DS Lab module) listing under the Meta Information tab of the DS Lab Model Runner component.

Accessing the same registered model under the Data Pipeline module

Please Note: The model based on an Algorithm script can be registered as an API service. You can refer to the Publish a Model as an API Service section for more details.