Evaluate Variable Influence for Predictions (Spatial Statistics Tools)

Summary

Evaluates how variables influence predictions from a spatial statistics model (.ssm) file. The tool creates partial dependence plots allowing you to see how each explanatory variable influenced predictions from the input model.

Learn more about how Evaluate Variable Influence for Predictions works

Illustration

Evaluate Variable Influence for Predictions tool illustration

Usage

The input model file must be created using the Forest-based and Boosted Classification and Regression tool.
You must provide input features that contain the same explanatory variables that were used to train the input model file, along with matching explanatory variable fields, distance features, and explanatory rasters. The tool creates an output table containing charts of the partial dependence plots for each explanatory variable when predicting to the input features. These charts display how each explanatory variable influenced the predicted values on average, along with the shape of the relationship.
When the dependent variable is categorical (a classification model), the partial dependence plots will show the average probability of classifying the first category. You can change the category displayed by the chart in the Chart Properties pane. In the Data tab, press the Select button under Numeric field(s), and choose the category to display. Only turn on one category at a time for the chart to correctly display the partial dependence plot.
The geoprocessing messages display various information about the input model file, including model parameters and training statistics. The contents of the messages depend on how the model was created.

Label	Explanation	Data type
Input Model File	The spatial statistics model file that will be used to make predictions and generate partial dependence plots. The model file must be created using the Forest-based and Boosted Classification and Regression tool.	File
Evaluation Type	Specifies the method that will be used to evaluate variable influence. Only partial dependence plots are currently supported. Partial dependence plots—Evaluates variable influence using partial dependence plots.	String
Input Features	The input features from which predictions will be evaluated. The input features must contain matching fields to the features that were used to create the input model file.	Feature Layer
Output Features (Optional)	The output features. This parameter is not currently used.	Feature Class
Output Table (Optional)	The output table containing the partial dependence plot values. The table includes charts that display the partial dependence plots of each explanatory variable.	Table
Match Explanatory Variables (Optional)	A list of the explanatory variables of the input model and corresponding fields of the input features. For each explanatory variable in the `Training` column, provide the corresponding prediction field in the `Prediction` column. The `Categorical` column specifies whether the variable is categorical or continuous. Value table columns: Prediction—The prediction field of the input features. Training—The explanatory variable of the input model file. Categorical—Specifies whether the variable is categorical or continuous. Categorical—The explanatory variable is categorical. Numeric—The explanatory variable is continuous.	Value Table
Match Distance Features (Optional)	A list of the explanatory distance features of the input model and corresponding prediction distance features. For each explanatory distance features in the `Training` column, provide the corresponding prediction distance features in the `Prediction` column. Value table columns: Prediction—The prediction distance features. Training—The explanatory distance features of the input model file.	Value Table
Match Explanatory Rasters (Optional)	A list of the explanatory rasters of the input model and corresponding prediction rasters. For each explanatory raster in the `Training` column, provide the corresponding prediction raster in the `Prediction` column. The `Categorical` column specifies whether the raster is categorical or continuous. Value table columns: Prediction—The prediction raster. Training—The explanatory raster of the input model file. Categorical—Specifies whether the raster is categorical or continuous. Categorical—The explanatory raster is categorical. Numeric—The explanatory raster is continuous.	Value Table

arcpy.stats.EvaluateVariableInfluence(input_model, evaluation_type, input_features, {output_features}, {out_table}, {explanatory_variable_matching}, {explanatory_distance_matching}, {explanatory_rasters_matching})

Name	Explanation	Data type
input_model	The spatial statistics model file that will be used to make predictions and generate partial dependence plots. The model file must be created using the Forest-based and Boosted Classification and Regression tool.	File
evaluation_type	Specifies the method that will be used to evaluate variable influence. Only partial dependence plots are currently supported. `PDP`—Evaluates variable influence using partial dependence plots.	String
input_features	The input features from which predictions will be evaluated. The input features must contain matching fields to the features that were used to create the input model file.	Feature Layer
output_features (Optional)	The output features. This parameter is not currently used.	Feature Class
out_table (Optional)	The output table containing the partial dependence plot values. The table includes charts that display the partial dependence plots of each explanatory variable.	Table
explanatory_variable_matching [[pred1, train1, cat1], [pred2, train2, cat2],...] (Optional)	A list of the explanatory variables of the input model and corresponding fields of the input features. For each explanatory variable in the `Training` column, provide the corresponding prediction field in the `Prediction` column. The `Categorical` column specifies whether the variable is categorical or continuous. Value table columns: `Prediction`—The prediction field of the input features. `Training`—The explanatory variable of the input model file. `Categorical`—Specifies whether the variable is categorical or continuous. `CATEGORICAL`—The explanatory variable is categorical. `NUMERIC`—The explanatory variable is continuous.	Value Table
explanatory_distance_matching [[pred1, cat1], [pred2, cat2],...] (Optional)	A list of the explanatory distance features of the input model and corresponding prediction distance features. For each explanatory distance features in the `Training` column, provide the corresponding prediction distance features in the `Prediction` column. Value table columns: `Prediction`—The prediction distance features. `Training`—The explanatory distance features of the input model file.	Value Table
explanatory_rasters_matching [[pred1, train1, cat1], [pred2, train2, cat2],...] (Optional)	A list of the explanatory rasters of the input model and corresponding prediction rasters. For each explanatory raster in the `Training` column, provide the corresponding prediction raster in the `Prediction` column. The `Categorical` column specifies whether the raster is categorical or continuous. Value table columns: `Prediction`—The prediction raster. `Training`—The explanatory raster of the input model file. `Categorical`—Specifies whether the raster is categorical or continuous. `CATEGORICAL`—The explanatory raster is categorical. `NUMERIC`—The explanatory raster is continuous.	Value Table

Code sample

EvaluateVariableInfluence example 1 (Python window)

The following Python window script demonstrates how to use the EvaluateVariableInfluence function.

# Create partial dependence plots
arcpy.stats.EvaluateVariableInfluence(
    input_model=r"C:\Data\My_Forest_Model.ssm",
    evaluation_type="PDP",
    input_features="MyLayer",
    out_table="PDP_Chart_Table",
    explanatory_variable_matching="CatVariable MATCHVAR1 true;ContVariable MATCHVAR2"
)

EvaluateVariableInfluence example 2 (stand-alone script)

The following stand-alone Python script demonstrates how to use the EvaluateVariableInfluence function.

# Create partial dependence plots.  

import arcpy 

# Set the current workspace.
arcpy.env.workspace = r"c:\mydata\mydata.gdb" 

# Run tool.
arcpy.stats.EvaluateVariableInfluence(
    input_model=r"C:\mydata\My_XGBoost_Model.ssm",
    evalutation_type="PDP",
    input_features="Minnesota",
    output_features=None,
    out_table=r"PDP_Chart_Table",
    explanatory_variable_matching="v1_value V1_VALUE false",
    explanatory_distance_matching="citycenter CITYCENTER",
    explanatory_rasters_matching="'LandCover.tif' 'LandCover.tif'"
)

Environments

This tool does not use any geoprocessing environments.

Licensing information

Basic: Limited
Spatial Analyst is required to use rasters.
Standard: Limited
Spatial Analyst is required to use rasters.
Advanced: Limited
Spatial Analyst is required to use rasters.