prea.util
Class Distance

java.lang.Object
  prea.util.Distance

public class Distance
extends java.lang.Object

This is a class implementing various distance measures of two vectors.

Since:

2012. 4. 20

Version:

1.1

Author:

Mingxuan Sun

Constructor Summary

Distance()

Method Summary

private static void	changePermutationIndex(int[] uItemID, int[] vItemID) Given two permutation, re-index the id to the set {1,2,.., n}, where n is the total items of the two permutations.
static void	computeAverageRank(double[] score, double[] prb) Return the average rank of each score with/without ties prb=(lowrank+(tie-1)/2)/(k+1)
static double	distanceKendall(int[] uItemID, double[] uScore, int[] vItemID, double[] vScore, int n) Return the Kendall's Tau distance for two rankings.
private static double	distanceKendallParsed(int[] uItemID, double[] uPrb, int[] vItemID, double[] vPrb, int n) Return intermediate Kendall's Tau distance for two rankings parsed by prb
static double	distanceNDCG(int[] uItemID, double[] relevance, int[] vItemID, double[] userScore) Return NDCG score for a ranked list given the scores and relevance of items in the list.
static double	distanceSpearman(int[] uItemID, double[] uScore, int[] vItemID, double[] vScore, int n) Return Spearman distance for two rankings.
static double	distanceSpearmanParsed(int[] uItemID, double[] uPrb, int[] vItemID, double[] vPrb, int n) Return the Spearman distance for two rankings parsed by probability
private static double	preferProbability(double avgRank_i, double avgRank_j) Return the probability of the item i is preferred to the item j.

Methods inherited from class java.lang.Object

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

Constructor Detail

Distance

public Distance()

Method Detail

distanceNDCG

public static double distanceNDCG(int[] uItemID,
                                  double[] relevance,
                                  int[] vItemID,
                                  double[] userScore)

Return NDCG score for a ranked list given the scores and relevance of items in the list.

Parameters:

uItemID - array of userIds

relevance - array of relevances for userIds

vItemID - array of userIds may/maynot be in the same order as uItemID

userScore - array of scores for userIds

Returns:

the NDCG score

changePermutationIndex

private static void changePermutationIndex(int[] uItemID,
                                           int[] vItemID)

Given two permutation, re-index the id to the set {1,2,.., n}, where n is the total items of the two permutations. The original itemId will be re-ordered.

Parameters:

uItemID - array of itemID sorted increasingly

vItemID - array of itemID sorted increasingly

distanceKendallParsed

private static double distanceKendallParsed(int[] uItemID,
                                            double[] uPrb,
                                            int[] vItemID,
                                            double[] vPrb,
                                            int n)

Return intermediate Kendall's Tau distance for two rankings parsed by prb

Parameters:

uItemID - array of itemID sorted increasingly

uPrb - array of probability for user u

vItemID - array of itemID for user v sorted increasingly

vPrb - array of probability for user v

n - the number of total items

Returns:

the Kendall's Tau scores

distanceKendall

public static double distanceKendall(int[] uItemID,
                                     double[] uScore,
                                     int[] vItemID,
                                     double[] vScore,
                                     int n)

Return the Kendall's Tau distance for two rankings.

Parameters:

uItemID - array of itemID

uScore - array of score for user u

vItemID - array of itemID for user v

vScore - array of scores for user v

n - the number of total items

Returns:

the Kendall's Tau distance

distanceSpearmanParsed

public static double distanceSpearmanParsed(int[] uItemID,
                                            double[] uPrb,
                                            int[] vItemID,
                                            double[] vPrb,
                                            int n)

Return the Spearman distance for two rankings parsed by probability

Parameters:

uItemID - array of itemID sorted increasingly

uPrb - array of probability for user u

vItemID - array of itemID for user v sorted increasingly

vPrb - array of probability for user v

n - the number of total items

Returns:

the Spearman scores

distanceSpearman

public static double distanceSpearman(int[] uItemID,
                                      double[] uScore,
                                      int[] vItemID,
                                      double[] vScore,
                                      int n)

Return Spearman distance for two rankings.

Parameters:

uItemID - array of itemID

uScore - array of score for user u

vItemID - array of itemID for user v

vScore - array of scores for user v

n - the number of total items

Returns:

the Spearman distance

computeAverageRank

public static void computeAverageRank(double[] score,
                                      double[] prb)

Return the average rank of each score with/without ties prb=(lowrank+(tie-1)/2)/(k+1)

Parameters:

score - The array of scores

prb - The array of average ranks will be filled.

preferProbability

private static double preferProbability(double avgRank_i,
                                        double avgRank_j)

Return the probability of the item i is preferred to the item j.

Parameters:

avgRank_i - The average rank of item i

avgRank_j - The average rank of item j

Returns:

the intermediate Kendall scores