-
Notifications
You must be signed in to change notification settings - Fork 42
/
SpearmanRank.java
174 lines (153 loc) · 5.05 KB
/
SpearmanRank.java
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
package uk.ac.cam.ha293.tweetlabel.eval;
import java.io.IOException;
import java.util.List;
import uk.ac.cam.ha293.tweetlabel.util.Tools;
import java.util.ArrayList;
import java.util.Collections;
import java.util.Map;
import java.util.HashMap;
import jsc.correlation.KendallCorrelation;
import jsc.correlation.LinearCorrelation;
import jsc.correlation.SpearmanCorrelation;
import jsc.datastructures.PairedData;
public class SpearmanRank {
public static double srcc(SimilarityMatrix sm1, SimilarityMatrix sm2) {
if(sm1.dimension() != sm2.dimension()) {
System.err.println("Similarity Matrices not of same dimension");
return -1.0;
}
List<Long> uids = Tools.getCSVUserIDs();
//create baseline similarity pairs
List<SimilarityPair> sps1 = new ArrayList<SimilarityPair>();
List<SimilarityPair> sps2 = new ArrayList<SimilarityPair>();
for(int uid1=0; uid1<uids.size(); uid1++) {
for(int uid2=uid1; uid2<uids.size();uid2++) {
SimilarityPair sp1 = new SimilarityPair(sm1,sm1.lookupID(uid1),sm2.lookupID(uid2));
SimilarityPair sp2 = new SimilarityPair(sm2,sm1.lookupID(uid1),sm2.lookupID(uid2));
if(sp1 == null || sp2 == null) {
//no categories for one of them
continue;
} else {
sps1.add(sp1);
sps2.add(sp2);
}
}
}
//sort both pairing lists
Collections.sort(sps1);
Collections.sort(sps2);
Collections.reverse(sps1);
Collections.reverse(sps2);
//For efficiency: need a map of pairs to ranks
//this makes it O(n), rather than O(n^2)
Map<SimilarityPair,Integer> rankMap = new HashMap<SimilarityPair,Integer>();
for(int i=0; i<sps1.size(); i++) {
//lose the 0-sims, they lend nothing
if(sps1.get(i).similarity()>0.0) {
rankMap.put(sps1.get(i),i);
}
}
double n = rankMap.keySet().size();
double rhoInc = 6.0 / (n*(n*n-1));
double rho = 1.0;
for(int i=0; i<sps2.size(); i++) {
//if(i%10000==0) System.out.println("i="+i+", rho="+rho);
if(rankMap.containsKey(sps2.get(i))) {
int diff = rankMap.get(sps2.get(i))-i;
rho -= rhoInc*diff*diff;
//System.out.println("Pair "+sps2.get(i).uid1()+","+sps2.get(i).uid2()+":"+sps1.get(i).similarity()+" at rank "+i+" appears at rank "+(diff+i)+" in other list");
}
}
return rho;
}
public static double jscSRCC(SimilarityMatrix sm1, SimilarityMatrix sm2) {
if(sm1.dimension() != sm2.dimension()) {
System.err.println("Similarity Matrices not of same dimension");
return -1.0;
}
List<Long> uids = Tools.getCSVUserIDs();
int numPairings = (sm1.dimension()*(sm1.dimension()+1)/2);
double[] pairedDataX = new double[numPairings];
double[] pairedDataY = new double[numPairings];
int pairingIndex = 0;
for(int i=0; i<uids.size(); i++) {
for(int j=i; j<uids.size(); j++) {
pairedDataX[pairingIndex] = sm1.getID(uids.get(i), uids.get(j));
pairedDataY[pairingIndex] = sm2.getID(uids.get(i), uids.get(j));
pairingIndex++;
}
}
PairedData pairedData = new PairedData(pairedDataX,pairedDataY);
//System.out.println("now running sc");
//SpearmanCorrelation sc = new SpearmanCorrelation(pairedData);
//return sc.getR();
return LinearCorrelation.correlationCoeff(pairedData);
/*
//create baseline similarity pairs
List<SimilarityPair> sps1 = new ArrayList<SimilarityPair>();
List<SimilarityPair> sps2 = new ArrayList<SimilarityPair>();
for(int uid1=0; uid1<uids.size(); uid1++) {
for(int uid2=uid1; uid2<uids.size();uid2++) {
SimilarityPair sp1 = new SimilarityPair(sm1,sm1.lookupID(uid1),sm2.lookupID(uid2));
SimilarityPair sp2 = new SimilarityPair(sm2,sm1.lookupID(uid1),sm2.lookupID(uid2));
if(sp1 == null || sp2 == null) {
//no categories for one of them
continue;
} else {
sps1.add(sp1);
sps2.add(sp2);
}
}
}
//sort both pairing lists
Collections.sort(sps1);
Collections.sort(sps2);
Collections.reverse(sps1);
Collections.reverse(sps2);
Map<SimilarityPair,Double> rankMap1 = new HashMap<SimilarityPair,Double>();
Map<SimilarityPair,Double> rankMap2 = new HashMap<SimilarityPair,Double>();
//rank first list
for(int i=0; i<sps1.size(); i++) {
double similarity = sps1.get(i).similarity();
//check for tied values
int j = i+1;
int count = 0;
while(j < sps1.size() && sps1.get(j).similarity()==similarity) {
count++;
j++;
}
if(count == 0) {
rankMap1.put(sps1.get(i),(double)i);
} else {
//work out the average rank, give it to all tied values
double newRank = (i+j)/2.0;
for(int k=i; k<j; k++) {
rankMap1.put(sps1.get(k), newRank);
}
i=j-1;
}
}
//rank second list
for(int i=0; i<sps2.size(); i++) {
double similarity = sps2.get(i).similarity();
//check for tied values
int j = i+1;
int count = 0;
while(j < sps2.size() && sps2.get(j).similarity()==similarity) {
count++;
j++;
}
if(count == 0) {
rankMap2.put(sps2.get(i),(double)i);
} else {
//work out the average rank, give it to all tied values
double newRank = (i+j)/2.0;
for(int k=i; k<j; k++) {
rankMap2.put(sps2.get(k), newRank);
}
i=j-1;
}
}
*/
}
}