Regression models
Regression predicts a continuous value for each configured numeric label attribute. Each attribute produces a continuous output layer.
Requirements
- A finalized dataset version with numeric common label attributes.
- Labels that contain values for the attributes to predict.
- A dataset that does not use reference spectra.
The model's Label attributes come from the dataset version and are read-only in the model settings. Update the dataset if the attributes are wrong or incomplete.
Evaluation
Regression results include RMSE, MAE, R², and SMAPE by attribute, plus predicted-versus-observed plots. Interpret each metric in the units and range of the target attribute; no single metric is sufficient for every application.