Package com.cobber.fta

Class Histogram

Object

Histogram

public class Histogram
extends Object

This class is used to encapsulate a Histogram to provide Histogram data. If the data fits in the cardinality set then it simply uses a map to generate the histogram values. Once the cardinality exceeds maxCardinality then the data is tracked using an algorithm based on Yael Ben-Haim and Elad Tom-Tov, "A streaming parallel decision tree algorithm", J. Machine Learning Research 11 (2010), pp. 849--872 All data is stored in the Cardinality Map until this is exhausted at this point we start to populate (via accept) the underlying Histogram Sketch with all values not captured in the Cardinality Map. Once we need to generate a Histogram we either just generate it from the Cardinality Map or if the MaxCardinality has been exceeded we add all the entries captured in the Cardinality Map to the Sketch.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
class	Histogram.Entry	A Histogram Entry captures the low and high bounds for each bucket along with the number of entries in the bucket.

Method Summary

Modifier and Type	Method	Description
static Histogram.Entry	getBucket(Histogram.Entry[] buckets, double value)	Given a value and set of buckets - locate the bucket holding this value.
Histogram.Entry[]	getHistogram(int buckets)	Get the histogram with the supplied number of buckets
Histogram	merge(Histogram other)
void	setCardinality(Map<String,Long> map)
void	setCardinalityOverflow(HistogramSPDT histogramSPDT)
static void	tagClusters(Histogram.Entry[] buckets)	Given a Histogram analysis mark each bucket as part of a cluster and then attach the count and percent for the cluster to all buckets in the cluster.

Methods inherited from class java.lang.Object

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

Method Detail

setCardinality

public void setCardinality(Map<String,Long> map)

setCardinalityOverflow

public void setCardinalityOverflow(HistogramSPDT histogramSPDT)

getHistogram

public Histogram.Entry[] getHistogram(int buckets)

Get the histogram with the supplied number of buckets

Parameters:

buckets - the number of buckets in the Histogram

Returns:

An array of length 'buckets' that constitutes the Histogram (or null if cardinality is zero).

merge

public Histogram merge(Histogram other)

tagClusters

public static void tagClusters(Histogram.Entry[] buckets)

Given a Histogram analysis mark each bucket as part of a cluster and then attach the count and percent for the cluster to all buckets in the cluster. For example, with the following distribution: 1, 1, 0, 0, 0, 0, 10, 20, 30, 30, 8 We would declare two clusters - the first one having 2% and the second having 98%, so the percentages would look as follows: 2, 2, 0, 0, 0, 0, 98, 98, 98, 98, 98

Parameters:

buckets - The set of Histogram buckets for this analysis

getBucket

public static Histogram.Entry getBucket(Histogram.Entry[] buckets,
                                        double value)

Given a value and set of buckets - locate the bucket holding this value.

Parameters:

buckets - The set of Histogram buckets for this analysis

value - The value we are searching for

Returns:

The bucket containing the supplied value.