Regression
Relative absolute error
Implemented
regression.raeDefinition
Total absolute error relative to that of always predicting the mean; below 1 beats the mean.
Formula
Σ|ŷᵢ − yᵢ| / Σ|yᵢ − ȳ|
Range: [0, ∞)
Inputs and outputs
- y_true: array of real targets
- y_pred: array of real predictions
- sample_weight: optional
Returns: MetricResult
Assumptions
No assumptions beyond valid, aligned inputs of the documented types.
Limitations
No metric-specific limitations are documented yet. Interpret the value alongside the task, data, and other metrics.
Python API
import evalsuite as es
es.rae(yr, pr)References
- Witten IH, Frank E, Hall MA. Data Mining. 3rd ed. Morgan Kaufmann; 2011.