hex.psvm

Class MetricBuilderPSVM<T extends MetricBuilderPSVM<T>>

java.lang.Object

water.Iced<T>

hex.ModelMetrics.MetricBuilder<T>

hex.ModelMetricsSupervised.MetricBuilderSupervised<T>

hex.psvm.MetricBuilderPSVM<T>

Type Parameters:

T -

All Implemented Interfaces:

java.io.Externalizable, java.io.Serializable, java.lang.Cloneable, water.Freezable<T>

public class MetricBuilderPSVM<T extends MetricBuilderPSVM<T>>
extends hex.ModelMetricsSupervised.MetricBuilderSupervised<T>

Binomial Metric builder tailored to SVM SVM doesn't predict probabilities, only probabilities 0-1 are returned, this renders some binomial metric misleading (eg. AUC, logloss) For maximum code re-use we still do use AUCBuilder, AUC2 instances provide confusion matrix and all metrics based on confusion matrix (accuracy, ...) This builder produces an instance of ModelMetricsBinomial, with irrelevant metrics undefined (NaN)

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected hex.AUC2.AUCBuilder	_auc

Fields inherited from class hex.ModelMetricsSupervised.MetricBuilderSupervised

_domain, _nclasses

Fields inherited from class hex.ModelMetrics.MetricBuilder

_CMetricScoringTask, _count, _customMetric, _sumsqe, _wcount, _work, _wY, _wYY

Constructor Summary

Constructors

Constructor and Description
MetricBuilderPSVM(java.lang.String[] domain)

Method Summary

Modifier and Type	Method and Description
hex.ModelMetrics	makeModelMetrics(hex.Model m, water.fvec.Frame f, water.fvec.Frame frameWithWeights, water.fvec.Frame preds) Create a ModelMetrics for a given model and frame
double[]	perRow(double[] ds, float[] yact, double w, double o, hex.Model m)
double[]	perRow(double[] ds, float[] yact, hex.Model m)
void	reduce(T mb)
java.lang.String	toString()

Methods inherited from class hex.ModelMetrics.MetricBuilder

cachePrediction, makePredictionCache, postGlobal, postGlobal, reduce, reduceForCV, setCustomMetric, weightedSigma

Methods inherited from class water.Iced

asBytes, clone, copyOver, frozenType, read, readExternal, readJSON, reloadFromBytes, toJsonBytes, toJsonString, write, writeExternal, writeJSON

Methods inherited from class java.lang.Object

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

Field Detail

_auc

protected hex.AUC2.AUCBuilder _auc

Constructor Detail

MetricBuilderPSVM

public MetricBuilderPSVM(java.lang.String[] domain)

Method Detail

perRow

public double[] perRow(double[] ds,
                       float[] yact,
                       hex.Model m)

Specified by:

perRow in class hex.ModelMetrics.MetricBuilder<T extends MetricBuilderPSVM<T>>

perRow

public double[] perRow(double[] ds,
                       float[] yact,
                       double w,
                       double o,
                       hex.Model m)

Overrides:

perRow in class hex.ModelMetrics.MetricBuilder<T extends MetricBuilderPSVM<T>>

reduce

public void reduce(T mb)

Overrides:

reduce in class hex.ModelMetrics.MetricBuilder<T extends MetricBuilderPSVM<T>>

makeModelMetrics

public hex.ModelMetrics makeModelMetrics(hex.Model m,
                                         water.fvec.Frame f,
                                         water.fvec.Frame frameWithWeights,
                                         water.fvec.Frame preds)

Create a ModelMetrics for a given model and frame

Specified by:

makeModelMetrics in class hex.ModelMetrics.MetricBuilder<T extends MetricBuilderPSVM<T>>

Parameters:

m - Model

f - Frame

frameWithWeights - Frame that contains extra columns such as weights

preds - Optional predictions (can be null), only used to compute Gains/Lift table for binomial problems @return

Returns:

ModelMetricsBinomial

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object