# Subspace and Correlation Clustering

I've compiled a list of ELKI invocations for subspace/correlation clustering with nicely fit parameters (except for PROCLUS). This post is actually more of a reminder for myself. A note on PROCLUS: Due to the randomness of k-medoids it seems I only get good clusters half of the time. It's something to consider.

SUBCLU
java -jar elki.jar KDDCLIApplication -algorithm clustering.subspace.SUBCLU -dbc.in exampledata2.txt -subclu.epsilon 3 -subclu.minpts 1 -subclu.mindim 1 -distance.dims 1,2,3

CLIQUE
java -jar elki.jar KDDCLIApplication -algorithm clustering.subspace.CLIQUE -dbc.in exampledata3.txt -clique.tau 0.3 -clique.xsi 8

PROCLUS
java -jar elki.jar KDDCLIApplication -algorithm clustering.subspace.PROCLUS -dbc.in exampledata3.txt -proclus.mi 10 -projectedclustering.k 2 -projectedclustering.l 2

PreDeCon
java -jar elki.jar KDDCLIApplication -algorithm clustering.subspace.PreDeCon -dbc.in exampledata5.txt -predecon.delta 0.25 -predecon.kappa 100 -predecon.lambda 2 -dbscan.minpts 3 -dbscan.epsilon 1

ORCLUS
java -jar elki.jar KDDCLIApplication -algorithm clustering.correlation.ORCLUS -dbc.in exampledata5.txt -orclus.alpha 0.5 -projectedclustering.k 2 -projectedclustering.l 1

4C
java -jar elki.jar KDDCLIApplication -algorithm clustering.correlation.FourC -dbc.in exampledata5.txt -dbscan.epsilon 1 -dbscan.minpts 3 -predecon.kappa 100 -pca.filter.delta 0.5

Leader
java -jar elki.jar KDDCLIApplication -algorithm clustering.Leader -dbc.in exampledata5.txt -leader.threshold 5

exampledata2.txt
2 5 2
1 5 1
9 5 9
8 5 9

exampledata3.txt
4.5, 4.0, 1.2
4.1, 2.0, 1.3
4.2, 3.0, 1.3
4.4, 2.0, 1.2
4.3, 5.0, 1.1
1.3, 4.0, 1.5
1.2, 5.0, 1.6
1.3, 2.0, 1.5
1.4, 3.0, 1.6

exampledata5.txt
1 4
2 4
3 4
4 4
5 4
6 4
7 1
7 2
7 3
7 4
7 5
7 6