raster_tools.general.model_predict_raster#

raster_tools.general.model_predict_raster(raster, model, n_outputs=1)[source]#

Generate a new raster using the provided model to predict new values.

The raster’s values are used as the predictor inputs for model. Each band in the input raster is used as a separate input variable. Outputs are raster surfaces where each band corresponds to a variable output by model.

The model argument must provide a predict method. If the desired model does not provide a predict function, ModelPredictAdaptor can be used to wrap it and make it compatible with this function.

Parameters

  • model (object) – The model used to estimate new values. It must have a predict method that takes an array-like object of shape (N, M), where N is the number of samples and M is the number of features/predictor variables. The predict method should return an (N, [n_outputs]) shape result. If only one variable is resurned, then the n_outputs dimension is optional.

  • n_outputs (int, optional) – The number of output variables from the model. Each output variable produced by the model is converted to a band in the output raster. The default is 1.

Returns

The resulting raster of estimated values. Each band corresponds to an output variable produced by model.

Return type

Raster