Skip navigation links

com.sigmadynamics.server.rtis.PredictionEngine
Interface PredictorInterface

public interface PredictorInterface

Used by ILS code to query Prediction Models.

Since:

1.0

Method Summary

void	flushLearningBuffer() Flushes learning buffers
int	getApplicationMappingId()
void	getExpectedValues(ModelInterface model, InputProvider input, ValueFilter[] outputFilter, SDDoubleArray values) Returns predicted numeric values for a set of values of partitioning attributes specified by an output filter.
void	getExpectedValues(ModelInterface model, InputProvider input, ValueFilter[] outputFilter, SDDoubleArray values, SDDoubleArray standardDeviations) Returns predicted numeric values and their standard deviations for a set of values of partitioning attributes specified by an output filter.
void	getLikelihoods(ModelInterface model, InputProvider input, ValueFilter[] outputFilter, SDDoubleArray likelihoods, boolean relaxed) Returns predicted likelihoods for a set of values of target and partitioning attributes specified by an output filter.
void	getLikelihoods(ModelInterface model, InputProvider input, ValueFilter[] outputFilter, SDDoubleArray likelihoods, SDDoubleArray standardDeviations, boolean relaxed) Returns predicted likelihoods for a set of values of target and partitioning attributes specified by an output filter.
int	getModelCount(ModelInterface model, ValueFilter[] outputFilter, ModelCount modelCount)
float	getModelError(ModelInterface model, ValueFilter[] outputFilter, ModelError modelError)
float	getModelLift(ModelInterface model, ValueFilter[] outputFilter, ModelLift modelLift)
float	getModelQuality(ModelInterface model, ValueFilter[] outputFilter)
void	getMostLikelyValues(ModelInterface model, InputProvider input, ValueFilter[] outputFilter, int outputAttributeId, int n, SDStringArray values, SDDoubleArray likelihoods) Returns predicted most likely values of a given attribute.
java.lang.String	getStudyName()
PredictionContribution[][]	getWhy(ModelInterface model, InputProvider input, ValueFilter[] outputFilter, int topN) Returns a list of attribute values that contribute the most to the model score.
void	learn(ModelInterface[] models, InputProvider inputProvider, LearningRecordWriterFactoryInterface learningRecordWriterFactory, long msTimestamp) Supplies learning information to the Prediction Engine.
void	learn(ModelInterface[] models, InputProvider inputProvider, long msTimestamp) Supplies learning information to the Prediction Engine.
void	learn(ModelInterface[] models, InputProvider inputProvider, SessionInterface session, long msTimestamp) Supplies learning information to the Prediction Engine.
void	loadModels() Loads new models if available.
ValueFilter[]	makeExplicitFilter(ModelInterface model, ValueFilter[] filter) Creates an explicit representation of a filter by filling in missing values, numeric intervals and missing sub-filters.

Method Detail

getLikelihoods

void getLikelihoods(ModelInterface model,
                    InputProvider input,
                    ValueFilter[] outputFilter,
                    SDDoubleArray likelihoods,
                    boolean relaxed)

Returns predicted likelihoods for a set of values of target and partitioning attributes specified by an output filter.

Parameters:

model - a model object in the application.

input - an input provider object.

outputFilter - a filter specifying what predictions are requested.

likelihoods - an array of predicted likelihoods.

relaxed - if this flag is true, then the model will return a likelihood prediction even if the prediction accuracy is very low. Otherwise, when prediction accuracy is low, the model will return Double.NaN.

getLikelihoods

void getLikelihoods(ModelInterface model,
                    InputProvider input,
                    ValueFilter[] outputFilter,
                    SDDoubleArray likelihoods,
                    SDDoubleArray standardDeviations,
                    boolean relaxed)

Returns predicted likelihoods for a set of values of target and partitioning attributes specified by an output filter.

Parameters:

model - a model object in the application.

input - an input provider object.

outputFilter - a filter specifying what predictions are requested.

likelihoods - output an array of predicted likelihoods.

standardDeviations - output an array of estimated standard deviations of likelihoods.

relaxed - if this flag is true, then the model will return a likelihood prediction even if the prediction accuracy is very low. Otherwise, when prediction accuracy is low, the model will return Double.NaN.

getWhy

PredictionContribution[][] getWhy(ModelInterface model,
                                  InputProvider input,
                                  ValueFilter[] outputFilter,
                                  int topN)

Returns a list of attribute values that contribute the most to the model score.

Parameters:

model - a model object in the application.

input - an input provider object.

outputFilter - a filter specifying what predictions are requested.

topN - number of top input score contributions to return.

getExpectedValues

void getExpectedValues(ModelInterface model,
                       InputProvider input,
                       ValueFilter[] outputFilter,
                       SDDoubleArray values)

Returns predicted numeric values for a set of values of partitioning attributes specified by an output filter.

Parameters:

model - a model object in the application.

input - an input provider object.

outputFilter - a filter specifying what predictions are requested.

values - an array of predicted expected values.

getExpectedValues

void getExpectedValues(ModelInterface model,
                       InputProvider input,
                       ValueFilter[] outputFilter,
                       SDDoubleArray values,
                       SDDoubleArray standardDeviations)

Returns predicted numeric values and their standard deviations for a set of values of partitioning attributes specified by an output filter.

Parameters:

model - a model object in the application.

input - an input provider object.

outputFilter - a filter specifying what predictions are requested.

values - an array of predicted expected values.

standardDeviations - an array of predicted standard deviations.

getMostLikelyValues

void getMostLikelyValues(ModelInterface model,
                         InputProvider input,
                         ValueFilter[] outputFilter,
                         int outputAttributeId,
                         int n,
                         SDStringArray values,
                         SDDoubleArray likelihoods)

Returns predicted most likely values of a given attribute.

Parameters:

model - a model object in the application.

input - an input provider object.

outputFilter - a filter specifying what predictions are requested.

outputAttributeId - ID of an attribute to return most likely values for.

n - a requested number of most likely values.

values - an array of returned most likely values.

makeExplicitFilter

ValueFilter[] makeExplicitFilter(ModelInterface model,
                                 ValueFilter[] filter)

Creates an explicit representation of a filter by filling in missing values, numeric intervals and missing sub-filters.

Parameters:

model - a model object in the application.

filter - a filter to make explicit. This object is not changed by the method.

Returns:

an explicit filter, possibly a reference to the argument filter object, if that filter is already explicit.

learn

void learn(ModelInterface[] models,
           InputProvider inputProvider,
           long msTimestamp)

Supplies learning information to the Prediction Engine.

Parameters:

models - an array of models that the learning applies to.

inputProvider - an input provider object responsible for providing values of input, partitioning and target attributes except the ones provided by the model object themselves.

msTimestamp - timestamp in milliseconds.

Since:

1.0

flushLearningBuffer

void flushLearningBuffer()

Flushes learning buffers

loadModels

void loadModels()

Loads new models if available. Waits until done.

getStudyName

java.lang.String getStudyName()

getApplicationMappingId

int getApplicationMappingId()

learn

void learn(ModelInterface[] models,
           InputProvider inputProvider,
           LearningRecordWriterFactoryInterface learningRecordWriterFactory,
           long msTimestamp)

Supplies learning information to the Prediction Engine.

Parameters:

models - an array of models that the learning applies to.

inputProvider - an input provider object responsible for providing values of input, partitioning and target attributes except the ones provided by the model object themselves.

learningRecordWriterFactory - used to write learning records either with autoFlush (for interactive sessions) or with manual flush (for batch sessions).

msTimestamp - timestamp in milliseconds.

learn

void learn(ModelInterface[] models,
           InputProvider inputProvider,
           SessionInterface session,
           long msTimestamp)

Supplies learning information to the Prediction Engine.

Called from generated code.

Parameters:

models - an array of models that the learning applies to.

inputProvider - an input provider object responsible for providing values of input, partitioning and target attributes except the ones provided by the model object themselves.

session - Contains a factory used to write learning records either directly to the learning table (for interactive sessions) or to the batch frameworks learning sandbox (for batch sessions).

msTimestamp - timestamp in milliseconds.

getModelCount

int getModelCount(ModelInterface model,
                  ValueFilter[] outputFilter,
                  ModelCount modelCount)

Returns:

counts in a model, a non-negative integer

See Also:

ChoiceModelInterface.getChoiceModelCount(ModelCount, String), ChoiceEventModelInterface.getChoiceEventModelCount(ModelCount, String, String)

getModelQuality

float getModelQuality(ModelInterface model,
                      ValueFilter[] outputFilter)

Returns:

quality for a model, a real number between 0 and 1 inclusive

See Also:

ChoiceModelInterface.getChoiceModelQuality(String), ChoiceEventModelInterface.getChoiceEventModelQuality(String, String)

getModelLift

float getModelLift(ModelInterface model,
                   ValueFilter[] outputFilter,
                   ModelLift modelLift)

Returns:

lift of a model, a real number of type float

See Also:

ChoiceModelInterface.getChoiceModelLift(ModelLift, String), ChoiceEventModelInterface.getChoiceEventModelLift(ModelLift, String, String)

getModelError

float getModelError(ModelInterface model,
                    ValueFilter[] outputFilter,
                    ModelError modelError)

Returns:

error of a model, a real number of type float

See Also:

ChoiceModelInterface.getChoiceModelError(ModelError, String), ChoiceEventModelInterface.getChoiceEventModelError(ModelError, String, String)

Skip navigation links