hex.genmodel.easy.prediction

Class BinomialModelPrediction

java.lang.Object

hex.genmodel.easy.prediction.AbstractPrediction

hex.genmodel.easy.prediction.BinomialModelPrediction

All Implemented Interfaces:

java.io.Serializable

public class BinomialModelPrediction
extends AbstractPrediction

Binomial classification model prediction. GLM logistic regression (GLM family "binomial") also falls into this category.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
double[]	calibratedClassProbabilities Class probabilities calibrated by Platt Scaling or Isotonic Regression.
double[]	classProbabilities This array of length two has the class probability for each class (aka categorical or factor level) in the response column.
float[]	contributions Per-feature prediction contributions (SHAP values).
java.lang.String	label Label of the predicted level.
int	labelIndex 0 or 1.
int[]	leafNodeAssignmentIds
java.lang.String[]	leafNodeAssignments
double[]	stageProbabilities Staged predictions of tree algorithms (prediction probabilities of trees per iteration).

Constructor Summary

Constructors

Constructor and Description
BinomialModelPrediction()

Method Summary

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

labelIndex

public int labelIndex

0 or 1.

label

public java.lang.String label

Label of the predicted level.

classProbabilities

public double[] classProbabilities

This array of length two has the class probability for each class (aka categorical or factor level) in the response column. The array corresponds to the level names returned by:

 model.getDomainValues(model.getResponseIdx())
 

"Domain" is the internal H2O term for level names. The values in this array may be Double.NaN, which means NA (this will happen with GLM, for example, if one of the input values for a new data point is NA). If they are valid numeric values, then they will sum up to 1.0.

calibratedClassProbabilities

public double[] calibratedClassProbabilities

Class probabilities calibrated by Platt Scaling or Isotonic Regression. Optional, only calculated if the model supports it.

leafNodeAssignments

public java.lang.String[] leafNodeAssignments

leafNodeAssignmentIds

public int[] leafNodeAssignmentIds

stageProbabilities

public double[] stageProbabilities

Staged predictions of tree algorithms (prediction probabilities of trees per iteration). The output structure is for tree Tt and class Cc: Binomial models: [probability T1.C1, probability T2.C1, ..., Tt.C1] where Tt.C1 correspond to the the probability p0 Multinomial models: [probability T1.C1, probability T1.C2, ..., Tt.Cc]

contributions

public float[] contributions

Per-feature prediction contributions (SHAP values). Size of the returned array is #features + 1 - there is a feature contribution column for each input feature, the last item is the model bias. The sum of the feature contributions and the bias term is equal to the raw prediction of the model. Raw prediction of tree-based model is the sum of the predictions of the individual trees before the inverse link function is applied to get the actual prediction. This means the sum is not equal to the probabilities returned in classProbabilities. (Optional) Available only for supported models (GBM, XGBoost).

Constructor Detail

BinomialModelPrediction

public BinomialModelPrediction()