C3 AI Documentation Home

C3 AI Studio Feature Store

The C3 AI Studio Feature Store is a centralized location for viewing features, feature sets, and related machine learning models in an application.

In the C3 Agentic AI Platform, a Feature represents a processed data point that captures a meaningful attribute or metric from a subject Type (for example, a wind turbine). These features serve as inputs to a related machine learning model. A Feature.Set combines multiple features into a single input for a machine learning model.

Data scientists and engineers use the Feature Store page to:

  • Collaborate and maintain traceability across the machine learning lifecycle.
  • Review how a feature was created (via code or JSON) to align team understanding or troubleshoot behavior.
  • Locate a feature in the store and include it in a new feature set for a different model.
  • Confirm the structure of a composite or Lambda feature set to ensure it meets the model’s training or inference needs.
  • Review the status of recent materialization jobs to identify errors or stale data.

Features

The Features tab displays a table of features created in the application. The table columns correspond to properties of the Feature Type. Feature Type properties include:

  • The feature name. Select this value to open a tabbed view with information about the feature.
  • The subject Type the feature belongs to
  • The kind of feature it is. Possible values include:
    • Metric: Metric features are created from C3 Metrics.

The header section displays the following feature metadata:

  • A summary of what the feature does (provided at its creation)
  • The kind of feature
  • The parent subject Type
  • Timestamps for feature creation, latest modification, and last materialization run

If the displayed feature is a time series, the time interval can be found here.

The Features tab contains two tabs:

Overview

The Overview tab contains:

  • Feature Definition: The code or JSON snippet used to create the feature.
  • Feature preview: A list of data that was materialized using the feature.

Learn about materializing data in Feature Materialization and Snapshots.

Associated models

The Associated models tab lists models that use, or were trained on, this feature.

Feature Sets

The Feature sets tab displays a table of feature sets in the application. The table columns show the following properties of the feature set:

  • Name
  • The kind of set. Possible values are:
    1. Lambda: A feature set created from a Python function.
    2. Composite: A feature set that is made by combining individual metric features.
  • The number of features in the feature set

Select a name from the table to open a tabbed view with information about the feature set. This header section shows metadata such as:

  • A brief description of the feature set, if one was provided at its creation
  • The associated subject type
  • The kind of feature set (Lambda or Composite)

Composite feature sets display a Features tab after the header, which lists the features used in the set. Each entry links to the individual features. Lambda feature sets display a Columns tab that shows the feature set Python source code, and a data column mapping that matches the function return value to a C3 AI Type.

All feature sets include the following tabs:

  • Associated models: A list of models that were trained on, or are currently using, the feature set.
  • Snapshots: A list of data that was cached from materialization runs.
  • Materialization runs: A list of previous runs and their statuses. Select a failed run to view its error details.

See also

Was this page helpful?