Predict Using AutoML (GeoAI Tools)

Summary

Predicts continuous variables (regression) or categorical variables (classification) on unseen compatible datasets using a trained .dlpk model produced by the Train Using AutoML tool.

Learn more about how AutoML works

Usage

You must install the proper deep learning framework for Python in ArcGIS Pro.

Learn how to install deep learning frameworks for ArcGIS
The input is an Esri model definition file (.emd) or a deep learning package file (.dlpk), which can be created using the Train Using AutoML tool.
A Spatial Analyst license is required to use rasters as explanatory variables or to predict to an Output Prediction Surface.
For information about requirements for running this tool and issues you may encounter, see Deep Learning frequently asked questions.

Label	Explanation	Data type
Model Definition	The `.dlpk` file or the `.emd` file.	File
Prediction Type	Specifies the output file type that will be created. Predict feature—A feature layer containing the prediction values will be created. The Output Prediction Features value is required for this option. Predict raster—A raster layer containing the prediction values will be created. The Output Prediction Surface value is required for this option.	String
Input Prediction Features	The features for which the prediction will be obtained. The input should include some or all the fields necessary to determine the dependent variable value. This parameter is required if the Prediction Type parameter is set to Predict feature.	Feature Layer; Table View; Feature Class
Explanatory Rasters (Optional)	A list of rasters that contain the explanatory rasters necessary to determine the dependent variable value. This parameter is required if the Prediction Type parameter is set to Predict raster.	Raster Layer
Distance Features (Optional)	The point or polygon features whose distances from the input training features will be estimated automatically and added as explanatory variables. Distances will be calculated from each of the input explanatory training distance features to the nearest input training features. If the input explanatory training distance features are polygons, the distance attributes will be calculated as the distance between the closest segments of the pair of features.	Feature Layer
Output Prediction Features	The output table or feature class.	Feature Class; Table
Output Prediction Surface	The path to where the output prediction raster will be saved.	Folder
Match Explanatory Variables (Optional)	The mapping of field names from the prediction set to the training set. Use this parameter if the field names of the training and prediction sets are different. The values are the field names in the prediction dataset that match the field names in the input feature class. Value table columns: Prediction—Field names in the prediction dataset. Training—Field names in the input feature class.	Value Table
Match Distance Variables (Optional)	The mapping of distance feature names from the prediction set to the training set. Use this parameter if the distance feature names used in the training and prediction sets are different. The string values are the feature names that were used for prediction that match the distance features names used for training. Value table columns: Prediction—The feature names that were used for prediction. Training—The distance features names that were used for training.	Value Table
Match Explanatory Rasters (Optional)	The mapping of names from the prediction rasters to the training rasters. Use this parameter if the names of the explanatory rasters used for prediction and the names of the corresponding rasters used during training are different. The string values are the explanatory raster names that were used for prediction that match the explanatory raster names used for training. Value table columns: Prediction—The explanatory raster names that were used for prediction. Training—The explanatory raster names used for training.	Value Table
Get explanation for every prediction (Optional)	Specifies whether fields representing feature importance will be added. Checked—Fields representing feature importance will be added. Fields will be generated for every predicted sample along with the prediction values. Unchecked—Fields representing feature importance will not be generated. This is the default.	Boolean

arcpy.geoai.PredictUsingAutoML(in_model_definition, prediction_type, in_features, {explanatory_rasters}, {distance_features}, out_prediction_features, out_prediction_surface, {match_explanatory_variables}, {match_distance_variables}, {match_explanatory_rasters}, {get_prediction_explanations})

Name	Explanation	Data type
in_model_definition	The `.dlpk` file or the `.emd` file.	File
prediction_type	Specifies the output file type that will be created. `PREDICT_FEATURE`—A feature layer containing the prediction values will be created. The Output Prediction Features value is required for this option. `PREDICT_RASTER`—A raster layer containing the prediction values will be created. The Output Prediction Surface value is required for this option.	String
in_features	The features for which the prediction will be obtained. The input should include some or all the fields necessary to determine the dependent variable value. This parameter is required if the `prediction_type` parameter is set to `PREDICT_FEATURE`.	Feature Layer; Table View; Feature Class
explanatory_rasters [explanatory_rasters,...] (Optional)	A list of rasters that contain the explanatory rasters necessary to determine the dependent variable value. This parameter is required if the `prediction_type` parameter is set to `PREDICT_RASTER`.	Raster Layer
distance_features [distance_features,...] (Optional)	The point or polygon features whose distances from the input training features will be estimated automatically and added as explanatory variables. Distances will be calculated from each of the input explanatory training distance features to the nearest input training features. If the input explanatory training distance features are polygons, the distance attributes will be calculated as the distance between the closest segments of the pair of features.	Feature Layer
out_prediction_features	The output table or feature class.	Feature Class; Table
out_prediction_surface	The path to where the output prediction raster will be saved.	Folder
match_explanatory_variables [match_explanatory_variables,...] (Optional)	The mapping of field names from the prediction set to the training set. Use this parameter if the field names of the training and prediction sets are different. The values are the field names in the prediction dataset that match the field names in the input feature class. Value table columns: `Prediction`—Field names in the prediction dataset. `Training`—Field names in the input feature class.	Value Table
match_distance_variables [match_distance_variables,...] (Optional)	The mapping of distance feature names from the prediction set to the training set. Use this parameter if the distance feature names used in the training and prediction sets are different. The string values are the feature names that were used for prediction that match the distance features names used for training. Value table columns: `Prediction`—The feature names that were used for prediction. `Training`—The distance features names that were used for training.	Value Table
match_explanatory_rasters [match_explanatory_rasters,...] (Optional)	The mapping of names from the prediction rasters to the training rasters. Use this parameter if the names of the explanatory rasters used for prediction and the names of the corresponding rasters used during training are different. The string values are the explanatory raster names that were used for prediction that match the explanatory raster names used for training. Value table columns: `Prediction`—The explanatory raster names that were used for prediction. `Training`—The explanatory raster names used for training.	Value Table
get_prediction_explanations (Optional)	Specifies whether fields representing feature importance will be added. `TRUE`—Fields representing feature importance will be added. Fields will be generated for every predicted sample along with the prediction values. `FALSE`—Fields representing feature importance will not be generated. This is the default.	Boolean

Code sample

PredictUsingAutoML example (stand-alone script)

This example shows how to use the PredictUsingAutoML function.

# Description: Predicts on feature or tabular data with the trained model
#              obtained by the TrainUsingAutoML function.

# Import system modules
import arcpy
import os

# Set local variables
datapath  = "path_to_data_for_prediction"
out_path = "path_to_gdb_for_predicted"

model_path = os.path.join(out_path, "model.dlpk")
in_features = os.path.join(datapath, "test_data.gdb", "test_data")
out_features = os.path.join(out_path, "predicted_feature.gdb", "predicted")

# Run Predict Using AutoML Model
r = arcpy.geoai.PredictUsingAutoML(model_path, "PREDICT_FEATURES",
                                   in_features,
                                   None, None, out_features, None, None, None, True)

Environments

Output Coordinate System, Geographic Transformations

Licensing information