Package moa.recommender.predictor

Class BRISMFPredictor

java.lang.Object

moa.AbstractMOAObject

moa.options.AbstractOptionHandler

moa.recommender.predictor.BRISMFPredictor

All Implemented Interfaces:

Configurable, java.io.Serializable, MOAObject, OptionHandler, RatingPredictor

public class BRISMFPredictor
extends AbstractOptionHandler
implements RatingPredictor

Implementation of the algorithm described in Scalable Collaborative Filtering Approaches for Large Recommender Systems (Gábor Takács, István Pilászy, Bottyán Németh, and Domonkos Tikk). A feature vector is learned for every user and item, so that the prediction of a rating is roughly the dot product of the corresponding user and item vector. Stochastic gradient descent is used to train the model, minimizing its prediction error. Both Tikhonov regularization and early stopping are used to reduce overfitting. The algorithm allows batch training (from scratch, using all ratings available at the moment) as well as incremental, by retraining only the affected user and item vectors when a new rating is inserted.

Parameters:

• f: features - the number of features to be trained for each user and item
• r: learning rate - the learning rate used in the regularization
• a: ratio - the regularization ratio to be used in the Tikhonov regularization
• i: iterations - the number of iterations to be used when retraining user and item features (online training).

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
ClassOption	dataOption
IntOption	featuresOption
IntOption	iterationsOption
FloatOption	lRateOption
FloatOption	rFactorOption
protected BRISMFPredictor	rp

Fields inherited from class moa.options.AbstractOptionHandler

config

Constructor Summary

Constructors

Constructor	Description
BRISMFPredictor()

Method Summary

Modifier and Type	Method	Description
RecommenderData	getData()
void	getDescription(java.lang.StringBuilder sb, int indent)	Returns a string representation of this object.
double	predictRating(int userID, int itemID)
double	predictRating(java.lang.Integer user, java.lang.Integer item)
java.util.List<java.lang.Double>	predictRatings(int userID, java.util.List<java.lang.Integer> itemIDS)
protected void	prepareForUseImpl(TaskMonitor monitor, ObjectRepository repository)	This method describes the implementation of how to prepare this object for use.
void	train()

Methods inherited from class moa.AbstractMOAObject

copy, measureByteSize, measureByteSize, toString

Methods inherited from class moa.options.AbstractOptionHandler

copy, getCLICreationString, getOptions, getPreparedClassOption, getPurposeString, prepareClassOptions, prepareForUse, prepareForUse

Methods inherited from interface moa.MOAObject

measureByteSize

Methods inherited from class java.lang.Object

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

Field Detail

rp

protected BRISMFPredictor rp

featuresOption

public IntOption featuresOption

lRateOption

public FloatOption lRateOption

rFactorOption

public FloatOption rFactorOption

iterationsOption

public IntOption iterationsOption

dataOption

public ClassOption dataOption

Constructor Detail

BRISMFPredictor

public BRISMFPredictor()

Method Detail

prepareForUseImpl

protected void prepareForUseImpl(TaskMonitor monitor,
                                 ObjectRepository repository)

Description copied from class: AbstractOptionHandler

This method describes the implementation of how to prepare this object for use. All classes that extends this class have to implement prepareForUseImpl and not prepareForUse since prepareForUse calls prepareForUseImpl.

Specified by:

prepareForUseImpl in class AbstractOptionHandler

Parameters:

monitor - the TaskMonitor to use

repository - the ObjectRepository to use

getDescription

public void getDescription(java.lang.StringBuilder sb,
                           int indent)

Description copied from interface: MOAObject

Returns a string representation of this object. Used in AbstractMOAObject.toString to give a string representation of the object.

Specified by:

getDescription in interface MOAObject

Parameters:

sb - the stringbuilder to add the description

indent - the number of characters to indent

predictRating

public double predictRating(java.lang.Integer user,
                            java.lang.Integer item)

getData

public RecommenderData getData()

Specified by:

getData in interface RatingPredictor

predictRating

public double predictRating(int userID,
                            int itemID)

Specified by:

predictRating in interface RatingPredictor

predictRatings

public java.util.List<java.lang.Double> predictRatings(int userID,
                                                       java.util.List<java.lang.Integer> itemIDS)

Specified by:

predictRatings in interface RatingPredictor

train

public void train()

Specified by:

train in interface RatingPredictor