gov.sandia.cognition.statistics

Class ChiSquaredSimilarity

java.lang.Object

gov.sandia.cognition.statistics.ChiSquaredSimilarity

@PublicationReference(author="Yao-Tsung Chen, Meng Chang Chen",
                      title="Using chi-square statistics to measure similarities for text categorization",
                      type=Journal,
                      year=2011,
                      url="http://www.sciencedirect.com/science/article/pii/S0957417410008961#")
public class ChiSquaredSimilarity
extends java.lang.Object

A class for computing the chi-squared similarity between two vectors. A chi- squared test requires frequency vectors, typically representing documents, so all values in the vectors will be computed as non-negative values. The test assumes one vector represents a document in a given category, and another vector which is being tested to see if it is (likely) from the same distribution as the original vector. Note that the test is symmetric, so the choice of which vector is the categorized vector and which one is the testing vector is somewhat arbitrary.

Since:

3.4.2

Author:

trbroun

Constructor Summary

Constructors

Constructor and Description
ChiSquaredSimilarity(Vector categorizedVector, Vector testingVector) Basic constructor.

Method Summary

Modifier and Type	Method and Description
double	compute() Computes the chi-squared statistic of the two vectors.
double	computeCumulativeProbabilityValue() Computes the chi-squared similarity statistic, then uses that to compute a cumulative probability.
Vector	getCategorizedVector() Basic getter for the categorized vector.
Vector	getTestVector() Basic getter for the testing vector.
void	setCategorizedVector(Vector newCategorizedVector) Basic setter for the categorized vector.
void	setTestVector(Vector newTestVector) Basic setter for the test vector.

Methods inherited from class java.lang.Object

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

Constructor Detail

ChiSquaredSimilarity

public ChiSquaredSimilarity(Vector categorizedVector,
                            Vector testingVector)

Basic constructor. Sets the variables in the intuitive manner.

Parameters:

categorizedVector - The vector from a known category.

testingVector - The vector which is being tested to see if it comes from the same category.

Method Detail

setCategorizedVector

public void setCategorizedVector(Vector newCategorizedVector)

Basic setter for the categorized vector.

Parameters:

newCategorizedVector -

setTestVector

public void setTestVector(Vector newTestVector)

Basic setter for the test vector.

Parameters:

newTestVector -

getCategorizedVector

public Vector getCategorizedVector()

Basic getter for the categorized vector.

Returns:

The categorized vector.

getTestVector

public Vector getTestVector()

Basic getter for the testing vector.

Returns:

The test vector.

compute

public double compute()

Computes the chi-squared statistic of the two vectors. This is a raw number and needs to be fed into a chi-squared distribution to compute a probability. Both vectors must be non-zero.

Returns:

The chi-squared statistic.

computeCumulativeProbabilityValue

public double computeCumulativeProbabilityValue()

Computes the chi-squared similarity statistic, then uses that to compute a cumulative probability. Returns the probability that a chi-squared statistic falls between 0 and the critical value (the computed chi-squared statistic for the two supplied vectors). Naturally, a large chi-squared value generates a large cumulative probability value.

Returns:

The probability of a chi-squared statistic being lower than the value of the chi-squared similarity of the given vectors.