Clustering using semEP

Year 2002
FC: 257
FI:  68		 Year 2003
FC: 407
FI:  78		 Year 2004
FC: 572
FI:  89		 Year 2005
FC: 657
FI:  90		 Year 2006
FC: 734
FI:  90		 Year 2007
FC: 445
FI:  83		 Year 2008
FC: 47
FI:  30		 All
Thresholds: 0.0 and 0.01
17 clusters 		Thresholds: 0.0 and 0.01
22 clusters 		Thresholds: 0.0 and 0.01
35 clusters 		Thresholds: 0.0 and 0.01
42 clusters 		Thresholds: 0.0 and 0.01
42 clusters 		Thresholds: 0.0 and 0.01
27 clusters 		Thresholds: 0.0 and 0.01
2 clusters 		About 300 clusters. Visualization is very heavy so that it hangs or is VERY SLOW.
Thresholds: 0.0 and 0.05
17 clusters 		Thresholds: 0.0 and 0.05
22 clusters 		Thresholds: 0.0 and 0.05
35 clusters 		Thresholds: 0.0 and 0.05
41 clusters 		Thresholds: 0.0 and 0.05
42 clusters 		Thresholds: 0.0 and 0.05
27 clusters 		Thresholds: 0.0 and 0.05
2 clusters
Thresholds: 0.0 and 0.1
2 clusters
Thresholds: 0.0 and 0.3
2 clusters
Thresholds: 0.0 and 0.7
2 clusters
Thresholds: 0.0 and 0.95
2 clusters


FC-FC similarity: either 0 or 1. Threshold=0 was used for the experiment.

FI-FI similarity: range [0 1]. Different thresholds were tried based on the similarity distribution but the pattern is the same -- the higher similarity, the more clusters.
Similarity histograms are here.