de.lmu.ifi.dbs.elki.algorithm.itemsetmining

Class APRIORI

java.lang.Object

de.lmu.ifi.dbs.elki.algorithm.AbstractAlgorithm<FrequentItemsetsResult>

de.lmu.ifi.dbs.elki.algorithm.itemsetmining.AbstractFrequentItemsetAlgorithm

de.lmu.ifi.dbs.elki.algorithm.itemsetmining.APRIORI

All Implemented Interfaces:

Algorithm

@Title(value="APRIORI: Algorithm for Mining Association Rules")
@Description(value="Searches for frequent itemsets")
@Reference(authors="R. Agrawal, R. Srikant",
           title="Fast Algorithms for Mining Association Rules",
           booktitle="Proc. 20th Int. Conf. on Very Large Data Bases (VLDB \'94), Santiago de Chile, Chile 1994",
           url="http://www.vldb.org/conf/1994/P487.PDF")
public class APRIORI
extends AbstractFrequentItemsetAlgorithm

The APRIORI algorithm for Mining Association Rules. Reference:

R. Agrawal, R. Srikant:
Fast Algorithms for Mining Association Rules
In Proc. 20th Int. Conf. on Very Large Data Bases (VLDB '94), Santiago de Chile, Chile 1994.

This implementation uses some simple optimizations for 1- and 2-itemsets. Note: this algorithm scales well to a large number of transactions, but not so well to a large number of frequent itemsets (items). For best results, use domain-specific preprocessing to aggregate items into groups. Use statistics logging to keep track of candidate set sizes.

Author:

Arthur Zimek, Erich Schubert

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	APRIORI.Parameterizer Parameterization class.

Field Summary

Fields

Modifier and Type	Field and Description
private static Logging	LOG The logger for this class.
private String	STAT Statistics logging prefix.

Fields inherited from class de.lmu.ifi.dbs.elki.algorithm.itemsetmining.AbstractFrequentItemsetAlgorithm

maxlength, minlength

Constructor Summary

Constructors

Constructor and Description
APRIORI(double minfreq) Constructor with minimum frequency.
APRIORI(double minfreq, int minlength, int maxlength) Constructor with minimum frequency.

Method Summary

Methods

Modifier and Type	Method and Description
protected List<Itemset>	aprioriGenerate(List<? extends Itemset> supported, int length, int dim) Prunes a given set of candidates to keep only those BitSets where all subsets of bits flipping one bit are frequent already.
private int	binarySearch(List<SparseItemset> candidates, SparseItemset scratch, int begin, int end) Binary-search for the next-larger element.
protected List<OneItemset>	buildFrequentOneItemsets(Relation<? extends SparseFeatureVector<?>> relation, int dim, int needed) Build the 1-itemsets.
protected List<SparseItemset>	buildFrequentTwoItemsets(List<OneItemset> oneitems, Relation<BitVector> relation, int dim, int needed, DBIDs ids, ArrayModifiableDBIDs survivors) Build the 2-itemsets.
private StringBuilder	debugDumpCandidates(StringBuilder msg, List<? extends Itemset> candidates, VectorFieldTypeInformation<BitVector> meta) Debug method: output all itemsets.
protected List<? extends Itemset>	frequentItemsets(List<? extends Itemset> candidates, Relation<BitVector> relation, int needed, DBIDs ids, ArrayModifiableDBIDs survivors, int length) Returns the frequent BitSets out of the given BitSets with respect to the given database.
protected List<SparseItemset>	frequentItemsetsSparse(List<SparseItemset> candidates, Relation<BitVector> relation, int needed, DBIDs ids, ArrayModifiableDBIDs survivors, int length) Returns the frequent BitSets out of the given BitSets with respect to the given database.
TypeInformation[]	getInputTypeRestriction() Get the input type restriction used for negotiating the data query.
protected Logging	getLogger() Get the (STATIC) logger for this class.
private boolean	initializeSearchItemset(BitVector bv, int[] scratchi, int[] iters) Initialize the scratch itemset.
private boolean	nextSearchItemset(BitVector bv, int[] scratchi, int[] iters) Advance scratch itemset to the next.
FrequentItemsetsResult	run(Relation<BitVector> relation) Performs the APRIORI algorithm on the given database.

Methods inherited from class de.lmu.ifi.dbs.elki.algorithm.itemsetmining.AbstractFrequentItemsetAlgorithm

getMinimumSupport

Methods inherited from class de.lmu.ifi.dbs.elki.algorithm.AbstractAlgorithm

makeParameterDistanceFunction, run

Methods inherited from class java.lang.Object

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

Field Detail

LOG

private static final Logging LOG

The logger for this class.

STAT

private final String STAT

Statistics logging prefix.

Constructor Detail

APRIORI

public APRIORI(double minfreq,
       int minlength,
       int maxlength)

Constructor with minimum frequency.

Parameters:

minfreq - Minimum frequency

minlength - Minimum length

maxlength - Maximum length

APRIORI

public APRIORI(double minfreq)

Constructor with minimum frequency.

Parameters:

minfreq - Minimum frequency

Method Detail

run

public FrequentItemsetsResult run(Relation<BitVector> relation)

Performs the APRIORI algorithm on the given database.

Parameters:

relation - the Relation to process

Returns:

the AprioriResult learned by this APRIORI

buildFrequentOneItemsets

protected List<OneItemset> buildFrequentOneItemsets(Relation<? extends SparseFeatureVector<?>> relation,
                                        int dim,
                                        int needed)

Build the 1-itemsets.

Parameters:

relation - Data relation

dim - Maximum dimensionality

needed - Minimum support needed

Returns:

1-itemsets

buildFrequentTwoItemsets

protected List<SparseItemset> buildFrequentTwoItemsets(List<OneItemset> oneitems,
                                           Relation<BitVector> relation,
                                           int dim,
                                           int needed,
                                           DBIDs ids,
                                           ArrayModifiableDBIDs survivors)

Build the 2-itemsets.

Parameters:

oneitems - Frequent 1-itemsets

relation - Data relation

dim - Maximum dimensionality

needed - Minimum support needed

ids - Objects to process

survivors - Output: objects that had at least two 1-frequent items.

Returns:

Frequent 2-itemsets

aprioriGenerate

protected List<Itemset> aprioriGenerate(List<? extends Itemset> supported,
                            int length,
                            int dim)

Prunes a given set of candidates to keep only those BitSets where all subsets of bits flipping one bit are frequent already.

Parameters:

supported - Support map

length - Itemset length

dim - Dimensionality

Returns:

itemsets that cannot be pruned by apriori

frequentItemsets

protected List<? extends Itemset> frequentItemsets(List<? extends Itemset> candidates,
                                       Relation<BitVector> relation,
                                       int needed,
                                       DBIDs ids,
                                       ArrayModifiableDBIDs survivors,
                                       int length)

Returns the frequent BitSets out of the given BitSets with respect to the given database.

Parameters:

candidates - the candidates to be evaluated

relation - the database to evaluate the candidates on

needed - Minimum support needed

ids - Objects to process

survivors - Output: objects that had at least two 1-frequent items.

length - Itemset length

Returns:

Itemsets with sufficient support

frequentItemsetsSparse

protected List<SparseItemset> frequentItemsetsSparse(List<SparseItemset> candidates,
                                         Relation<BitVector> relation,
                                         int needed,
                                         DBIDs ids,
                                         ArrayModifiableDBIDs survivors,
                                         int length)

Returns the frequent BitSets out of the given BitSets with respect to the given database. Optimized implementation for SparseItemset.

Parameters:

candidates - the candidates to be evaluated

relation - the database to evaluate the candidates on

needed - Minimum support needed

ids - Objects to process

survivors - Output: objects that had at least two 1-frequent items.

length - Itemset length

Returns:

Itemsets with sufficient support

initializeSearchItemset

private boolean initializeSearchItemset(BitVector bv,
                              int[] scratchi,
                              int[] iters)

Initialize the scratch itemset.

Parameters:

bv - Bit vector data source

scratchi - Scratch itemset

iters - Iterator array

Returns:

true if the itemset had minimum length

nextSearchItemset

private boolean nextSearchItemset(BitVector bv,
                        int[] scratchi,
                        int[] iters)

Advance scratch itemset to the next.

Parameters:

bv - Bit vector data source

scratchi - Scratch itemset

iters - Iterator array

Returns:

true if the itemset had minimum length

binarySearch

private int binarySearch(List<SparseItemset> candidates,
               SparseItemset scratch,
               int begin,
               int end)

Binary-search for the next-larger element.

Parameters:

candidates - Candidates to search for

scratch - Scratch space

begin - Search interval begin

end - Search interval end

Returns:

Position of first equal-or-larger element

debugDumpCandidates

private StringBuilder debugDumpCandidates(StringBuilder msg,
                                List<? extends Itemset> candidates,
                                VectorFieldTypeInformation<BitVector> meta)

Debug method: output all itemsets.

Parameters:

msg - Output buffer

candidates - Itemsets to dump

meta - Metadata for item labels

Returns:

Output buffer

getInputTypeRestriction

public TypeInformation[] getInputTypeRestriction()

Description copied from class: AbstractAlgorithm

Get the input type restriction used for negotiating the data query.

Specified by:

getInputTypeRestriction in interface Algorithm

Specified by:

getInputTypeRestriction in class AbstractAlgorithm<FrequentItemsetsResult>

Returns:

Type restriction

getLogger

protected Logging getLogger()

Description copied from class: AbstractAlgorithm

Get the (STATIC) logger for this class.

Specified by:

getLogger in class AbstractAlgorithm<FrequentItemsetsResult>

Returns:

the static logger