Regression evaluation
The Regression evaluation view returns performance metrics for a deployed regression model.
Use it to evaluate prediction error, compare evaluation data sources, and review summary metrics such as mean squared error, mean absolute error, mean absolute percentage error, R² score, and maximum error.
When to use
Generate the view
Generate the Regression evaluation view from a deployed model.
The request needs view_type set to REGRESSION_EVALUATION.
You can evaluate either a submitted query batch with target observations, or a precomputed observation split using observation_type.
When evaluating query data, provide both query and targets. When evaluating precomputed observations, set observation_type to TRAIN, TEST, or ALL.
Output
The response follows the shared Response structure format.
The Regression evaluation view returns a single dataframe in view_data.
Interpret the result
The Regression evaluation view is easiest to read by starting with the main error metrics, then using distribution fields to understand how consistent those errors are.
Start with absolute error
Use the mean_absolute_error_* fields to understand the typical absolute difference between predicted and observed values.
Mean absolute error is often the easiest metric to communicate because it is expressed in the same unit as the target.
Check squared error
Use the mean_squared_error_* fields when larger errors should be penalized more heavily.
Squared error grows quickly for large misses, so it is useful when outliers or large prediction errors are especially important.
Review percentage error carefully
Use the mean_absolute_percentage_error_* fields when relative error is useful.
Percentage error can become very large when actual target values are zero or close to zero, so interpret these fields in the context of the target distribution.
Inspect fit and worst-case error
Use r2_score to understand how much target variance is explained by the model.
Use max_error to identify the largest absolute error. This is useful for worst-case review, even when average errors look acceptable.
Compare evaluation sources
Use observation_type when you want to evaluate train, test, or all precomputed observations.
Comparing TRAIN and TEST results can help identify overfitting or performance drift between the data used to train the model and the data used to evaluate it.
Options
The Regression evaluation view supports one view-specific option.
Observation type
Use observation_type when you want the evaluation to run against precomputed observations.
When observation_type is not provided, the view evaluates the query and target data supplied in the request.
Field reference
The Regression evaluation view returns a dataframe, so individual columns are not documented as standalone API schema properties.
Squared error fields
These fields describe squared prediction error.
Absolute error fields
These fields describe absolute prediction error.
Percentage error fields
These fields describe absolute percentage prediction error.
Fit and maximum error fields
These fields describe fit and worst-case error.