Pre-configured dashboard templates enable teams to quickly view the health of their models, while customized dashboards allow for ad hoc analysis. Using dashboard visualizations of statistical distributions and performance heatmaps help focus teams on troubleshooting efforts.
Creating a Dashboard
The dashboards in Arize are designed to allow teams to deeply analyze and troubleshoot their models. We offer templates that make it quick and easy to launch dashboards for every model you want to monitor.
There are two main ways to create dashboards in the system:
Use a Template: This is a great starting point if you want a general visualization of the health of your models - most users start with dashboard templates, which they can then customize to suit their specific needs.
Start from a Blank Dashboard: Alternatively, you can build dashboards from scratch if you already have a good sense of the model dimensions you want to monitor and the type of analysis you wish to perform
Dashboards are comprised of widgets designed for different types of analysis across your training, validation, and production environments:
Distribution Widget for analyzing data distribution changes over Feature, Prediction, and Actuals.
Click the Edit Dashboard icon in the top right corner.
Select or drag and drop the widget onto an area of the dashboard.
You can easily change the size of each widget once it's on the dashboard, and you can drag and drop to new areas of the dashboard.
Slicing and Filtering Dashboards
You can slice and filter dashboards by any model, model version, and model dimension, including feature, prediction score/class, or actual score/class.
Compare Across Model Environments
You can set up a widget to compare across model environments (e.g. Production vs. Validation).
Comparing Production vs. Validation Data in Dashboard Widget
To set up a production vs. validation distribution graph in your dashboard:
1: Create a Dashboard > Click Edit Dashboard > Add a New Widget to the Dashboard, and select 'Distribution Widget'
2: Set up Production Distribution for this feature in Plot 1. Production data in the dashboard is displayed relative to the date picker selection. The date picker always adjusts production data.
3: Set up Validation Distribution for this feature in Plot 2. The model environment drop down for the validation data allows selection of the batch name. We support multiple batches for the validation data. Once the batch is chosen, the platform automatically shows the date for that batch of data. This fixed batch of data can then be compared against production data of any time frame.