hex.genmodel.descriptor

Interface ModelDescriptor

All Known Implementing Classes:

ModelDescriptorBuilder.MojoModelDescriptor, ModelDescriptorBuilder.PojoModelDescriptor

public interface ModelDescriptor

Internal structure providing basic information about a model. Used primarily, but no exclusively by MOJO pipeline. Every MOJO is guaranteed to provide the information defined by this interface.

Method Summary

Modifier and Type	Method and Description
java.lang.String	algoFullName()
java.lang.String	algoName()
boolean	balanceClasses()
java.lang.String[]	columnNames()
double	defaultThreshold() Default threshold for assigning class labels to the target class.
java.lang.String[]	features()
java.lang.String	foldColumn()
ModelCategory	getModelCategory() Model's category.
java.lang.String[][]	getOrigDomains()
java.lang.String[]	getOrigNames()
boolean	isSupervised()
double[]	modelClassDist()
int	nclasses()
int	nfeatures()
java.lang.String	offsetColumn()
double[]	priorClassDist()
java.lang.String	projectVersion() E.g.
java.lang.String[][]	scoringDomains() Domains of categorical features.
java.lang.String	timestamp()
java.lang.String	treatmentColumn()
java.lang.String	uuid()
java.lang.String	weightsColumn()

Method Detail

scoringDomains

java.lang.String[][] scoringDomains()

Domains of categorical features. For each feature, there is a record. If the feature is not categorical, the value is null.

Returns:

An array of String representing the complete domain of each feature. Null if the feature is not categorical.

projectVersion

java.lang.String projectVersion()

E.g. "3.24.0.1"

Returns:

A String representing version of H2O Open Source Machine Learning platform project.

algoName

java.lang.String algoName()

Returns:

A string with human-readable shortcut of the algorithm enveloped by this MOJO. Never null.

algoFullName

java.lang.String algoFullName()

Returns:

A string with human-readable, full textual representation of the algorithm. Never null.

offsetColumn

java.lang.String offsetColumn()

Returns:

A String with the name of the offset column used. Null if there was no offset column used during training.

weightsColumn

java.lang.String weightsColumn()

Returns:

A String with the name of the weights column used. Null if there were no weights applied to the dataset.

foldColumn

java.lang.String foldColumn()

Returns:

A String with the name of the fold column used. Null of there was no folding by using a fold column as a key done.

treatmentColumn

java.lang.String treatmentColumn()

Returns:

A String with the name of the treatment column used. Null of there was no treatment used during training.

getModelCategory

ModelCategory getModelCategory()

Model's category.

Returns:

One of ModelCategory values. Never null.

isSupervised

boolean isSupervised()

Returns:

True for supervised learning models, false for unsupervised.

nfeatures

int nfeatures()

Returns:

An integer representing a total count of features used for training of the model.

features

java.lang.String[] features()

Returns:

An array String representing the names of the features used for model training.

nclasses

int nclasses()

Returns:

Domain cardinality of the response column, Only meaningful if the model has a categorical response and the model is supervised.

columnNames

java.lang.String[] columnNames()

Returns:

An array of String representing the column names in the model. The very last one is the response column name, if the model has a response column.

balanceClasses

boolean balanceClasses()

defaultThreshold

double defaultThreshold()

Default threshold for assigning class labels to the target class. Applicable to binomial models only.

Returns:

A double primitive type with the default threshold value

priorClassDist

double[] priorClassDist()

modelClassDist

double[] modelClassDist()

uuid

java.lang.String uuid()

timestamp

java.lang.String timestamp()

getOrigNames

java.lang.String[] getOrigNames()

getOrigDomains

java.lang.String[][] getOrigDomains()