This page covers on how to manage and use dataapps.

DataApps

Dataapps are visualizations built using the react for four different problem types in Rapid model recipes such as Binary classification, Regression, multiclass classification, and Binary Experimental.

On the other hand, the DataApps can also be built by writing Python code on Juypter Notebook. If the DataApps are built from Notebook, business users can consume those once are deployed by data scientists. The data scientist will first create an app template on the Juypter Notebook and uses this app template to deploy the dataapp passing a project ID, scenario, artifact, or model. Business users can only view and delete the dataapps published onto the UI from the Notebook, but cannot create, modify the app templates. All the app templates created in a tenant are saved in the Data app module.

If you want to generate a chart, you can either create a custom template on the Notebook or use an existing system template. Examples of system templates you can use to generate the chart after building the machine learning flows are - Correlation Matrix, EDA data drift, Pivot Table viewer, Visual Analytics, Pandas Profiler, and so on. Using a system or custom template, you can generate the charts only with the default Streamlit version. Contrary to this, dataapps support all Streamlit versions.

The app template is similar to system templates, but app templates can only be used to generate dataapps(charts).

Accessing and Creating a DataApp at a project level

Use this procedure to create a DataApp in a project. After you run the data pipeline for any of these four problem types such as binary classification, regression, multiclass classification, or binary experimental problems, the DataApp button gets enabled.

1. Click the menu icon and select Projects

2. Select the project for which you want to create DataApps and click the DataApps icon from the project level navigation.

3. Click the plus icon or click +New DataApp button to create the DataApp. You can only view the later button when there are no dataapps created in this project.

The Create DataApp window appears.

4. Specify this information:

DataApp Name:

• The name of the DataApp.

DataApp Description:

• The description of the DataApp.

Recipe Name:

• The recipe for which you want to create a DataApp.

Choose Image:

• Click to upload an image to display on the DataApp card that you see on the DataApps page.

5. Click Create.

You can see the DataApp card created in this project.

Viewing dataapps in a tenant

Use this procedure to view the dataapps created on the UI or published by Notebook users from Notebook.

1. Click the menu icon and select DataApps. This displays the dataapps dashboard where you can see all the DataApps for different projects in this tenant.

Note

The DataApps will become inactive after a certain period set by the admin. If you want to use DataApps, you must relaunch.

2. Click Relaunch on the DataApp card that you want to launch again.

When the DataApp is running, the card appears as shown in this screenshot.

3. Click on the Dataapp. The screen displays the following tabs:

Feature importance - Assigning scores to input features based on how helpful they are in predicting a target variable. These scores are ranked to help users understand which features have a significant impact on model prediction.

Model performance - This evaluates the performance of the model.

What-If analysis - Help data scientists to get insights into what a model predicts for given input values. This allows them to experiment with various combinations of values for key features and observe the resulting predictions.

You can use the sliders to change the values or use drop-down to select values and click Predict What-If Outcome to review what the model predicts for the given input values.

Prediction - You can perform predictions on the uploaded dataset.

Performing predictions on the uploaded dataset

Use this procedure to do the predictions on the new dataset and generate charts using Ask AI.

1. Click Browse to upload the dataset and obtain predictions generated by the model.

   

2. Click Predict. Once the prediction is done on the dataset, you can download the prediction results, using the Download Prediction Results option.

3. Generate charts on the dataset generated after predictions, using the Ask AI recipe. This takes you to the Ask AI tab where you can provide the query in the query box and click the Generate button

Important

Apart from generating charts, you can also perform data transformation operations on the dataset by prompting for the Ask AI.

Here is the output chart generated for the given query.

4. Click +Add to DataApp to add this chart to the dataapp. The Update Name box appears where you can provide the custom name for the chart and click Save.

5. Click the Additional Charts tab to view this chart.

Ask AI - You can use the Ask AI feature similar to what you see in AI-assisted recipe to provide text prompt in the chat window and generate the chart outputs on the dataset used for model building. This allows you to generate visualizations and carry out different data pre-processing operations on the selected dataset.

You can add as many chats as you want and switch between the chat windows from the Chat list on the left. The difference between the chat window used for prediction output dataset and dataset used for model building can be see in its chat name on the left.

Perform the following actions clicking the caret icon:

• Copy the dataapps URL to share with the other business users, using the Copy option.

• Open the dataapps on a new tab, using the Open in New Tab option.

• Configure the hardware required to run dataapps, using the Config option.

• View logs of DataApps to debug issues, using the Logs option.

• Delete the DataApp that is no longer required, using the Delete option.

Dataapps dashboard

Dataapps dashboard has all the dataapps in the form of cards. These datapps are either created in projects or published from the Notebook. Each card gives various details such as project name, description of the project, recipe used, the person who created the dataapp, last shutdown by, and the date and time at which the dataapp was created. You can also launch the app and relaunch when the app is idle.

Use the search box to search for Dataapp templates by name.

The list view provides the Data app details in the table form. You can view the same details that you see in the tile view of Dataapps such as DataApp name, Created on, Created by, Description and Project.

Use the table settings icon to rearrange the columns in the order you want and view or hide the columns by selecting and clearing the check boxes next to the columns.

The card view provides all the data app cards on the dashboard. Click on each card to view the DataApp associated with the project.

Use the filter option to filer the dataapps by project and view only the dataapps associated with the filtered project on the dataapps dashboard.

1. Select or clear the projects whose dataapps you want to view or hide.

2. Click Apply.

Editing a DataApp

Use this procedure to delete a DataApp.

To edit a DataApp:

1. Click the menu icon and select DataAapps. This shows the DataApps created under this tenant and in different projects.

2. Click on the ellipses icon on a DataApp that you want to edit and select the Edit option. This opens the side panel window to make changes to DataApp details.

3. Modify the details and Click Save to update the changes.

Deleting a DataApp

Use this procedure to delete a DataApp.

To delete a DataApp:

1. From the left navigation menu, select DataAapps. The DataApps dashboard is displayed.

2. Click the ellipses icon on the DataApps dashboard.

3. Click Delete. The dialog box appears prompting you to delete or discard

4. Click Yes, Delete to delete the DataApp permanently.

Viewing DataApp logs

Use the history of DataApp logs to debug the issues in DataApp.

To view DataApp logs:

1. From the left navigation menu, select DataAapps. The DataApps dashboard is displayed.

2. Click the ellipses icon on the DataApps dashboard.

3. Click Logs. The dialog box appears prompting you to delete or discard.

You can now view the DataApp logs.

Configuring hardware to run dataapps

Use this procedure to configure the hardware required to run dataapps without any interruption. You can check the resource consumption by DataApps while the DataApps are running.

To view DataApp logs:

1. From the left navigation menu, select DataAapps. The DataApps dashboard is displayed.

2. Click the ellipses icon on the DataApps dashboard.

3. Click Configs.

The Configuration & Consumption pull-out window appears.

4. Select the environment in which you want to run the dataapps. Possible values:

   • EXTRA SMALL: 1 Core, 1GB Memory, and 5GB Disk Space

   • SMALL: 1 Core, 2GB Memory, and 20GB Disk Space

   • MEDIUM: 2 Cores, 4GB Memory, and 20GB Disk Space

   • LARGE: 4 Cores, 8GB Memory, and 20GB Disk Space

   • CPU_LARGE: 8 Cores, 16GB Memory, and 20GB Disk Space

   • MAX_LARGE: 12 Cores, 32GB Memory, and 20GB Disk Space

   • EXTRA_MAX_LARGE: 12 Cores, 48GB Memory, and 20GB Disk Space

Note

Dataapps and projects are run in different environments. You have to configure the environment for project and dataapps separately.

5. Specify the time after which the DataApp must shut down if there is no activity. You can set the time in hours.

6. Click Save.