what is the difference between manhattan and euclidean distances calculations on single linkage cluster ?? which one is preferable ? i wrote a program for single linkage clustering and used both distances calculations ,, i just see that when i want to divide clusters like 5 clusters or 6 clusters by stopping iterations ,, i seen clusters has different members at the same iterations.
like this :
for euclid
.....İTERATİON ..... :39
Cluster 0
1 7 19 33 46 47 20 27
Cluster 1
2 8 28 34 44 17 31 5 25
Cluster 2
3 11 12 37 39 29 42 15
Cluster 3
4 6 14 32 41 30
Cluster 8
9 16 35 43
Cluster 9
10 21 36
Cluster 12
13 26 40 18 22 45 24 38
Cluster 22
23
and for manhattan :
.....İTERATİON ..... :39
Cluster 0
1 7 20 33 47 46 19 27
Cluster 1
2 8 28 34 44 17 31 5 25
Cluster 2
3 11 12 37 39 29 42 15
Cluster 3
4 6 14 32 41 30
Cluster 8
9 16 35 43
Cluster 9
10 21 36
Cluster 12
13 26 40 18 22 45 24 38
Cluster 22
23