Classify Using Machine Learning

ERDAS IMAGINE Help

HGD_Variant
16.5
HGD_Product
ERDAS IMAGINE
HGD_Portfolio_Suite
Producer

Category: Classification

Associated Namespace: IMAGINE

Default

Show All Ports

ClassifyUsingMachineLearning_Default

ClassifyUsingMachineLearning_AllPorts

Description

This operator performs classification on the input data using the trained classifier specified on the MachineIntellect port.

The input data can be of type IMAGINE.Features or IMAGINE.Raster.

For feature input data, its feature schema must contain attribute fields whose names match with the non-geometry attribute fields of the training data used for training the MachineIntellect. The result is the same as the input feature data with additional attribute fields containing results from the operator. There will be one attribute field that contains the classes that the features are classified into. This field is named the same as the attribute field which was used for defining the classes when training the MachineIntellect. There will also be several additional fields (as many as the number of classes the data is classified into) that give the computed probability of the features being classified to each class if the ComputeProbabilities port is set to True.

For raster input data, the number of bands in the input raster must be equal to the number of Attributes used when training the MachineIntellect and they must be in the same order. The Band Selection and Stack Layers operators may be used to order the bands appropriately. The result of the classification is a single band thematic raster with pixel values representing the class into which a location is classified. The band is named the same as the attribute field which was used for defining the classes when training the MachineIntellect. The output raster will also have a Class_Names field in its attribute table that lists the classes.

Limitations

None

Connections

Name

Objects Supported

Description

Shown by Default

Required

Default Behavior or Behavior if not Required

DataIn

IMAGINE.Features

IMAGINE.Raster

The input data to be classified.

checkmark_gray

checkmark_gray

MachineIntellect

IMAGINE.MachineIntellect

A trained Machine Intellect, such as those produced by the Initialize CART, Initialize SVM, Initialize Random Forest, etc., operators.

checkmark_gray

checkmark_gray

ComputeProbabilities

IMAGINE.Bool

Flag to compute probability of the records in DataIn being classified to each class present in the model. The probability is added as an attribute (double precision floating point) corresponding to each class.

Probabilities are computed when DataIn is of IMAGINE.Features type.

False

ProbabilityAttributeNamePrefix

IMAGINE.String

The prefix for the output probability attributes.

The attribute name is prefixed with ProbabilityAttributeNamePrefix and suffixed with the class value.

A new double precision floating point field will be added for each class if it does not yet exist. If the attribute name matches the name of an existing field, the existing field will be overwritten.

Probability values are generated when DataIn is of IMAGINE.Features type and ComputeProbability is set to True.

CLP_

DataOut

IMAGINE.Features

IMAGINE.Raster

The classified data.

The output data type is based on the input data type. For Feature output, the classification result is added as an attribute(s) to the feature.

For Raster output, the output is a single band thematic raster representing the class a pixel is classified into. The attribute table of the raster includes a Class_Names field that lists the new classes.

checkmark_gray

Related Operators

Initialize CART, Initialize K-Nearest Neighbors, Initialize Naive Bayes, Initialize SVM, Initialize Random Forest

Example Model

InitializeSVM_example