public Dendrogram cluster() {
Dendrogram dnd = new Dendrogram("Distance");
double d = 0.0;
// initially load all elements as individual clusters
for (DataPoint e : elements) {
Cluster c = new Cluster(e);
allClusters.add(c);
}
dnd.addLevel(String.valueOf(d), allClusters.getAllClusters());
d = 1.0;
while (allClusters.size() > 1) {
int K = allClusters.size();
mergeClusters(d);
// it is possible that there were no clusters to merge for current
// d.
if (K > allClusters.size()) {
dnd.addLevel(String.valueOf(d), allClusters.getAllClusters());
K = allClusters.size();
}
d = d + 0.5;
}
return dnd;
}
/**
* Implements the core algorithm.
*
* @return
*/
protected Dendrogram getClusteringDendrogram() {
// 1. inizializza la matrice di prossimità con un cluster per ogni documento
ProximityMatrix m = this.initMatrix();
Dendrogram dendrogram = this.initDendrogram();
for (int i = 0; i < m.getClustersCount() - 1; i++) {
// 2. cerca la coppia di cluster più vicini (single link - distanza punti più vicini)
ICluster[] nearestClusters = m.getNearestClusters();
// TODO DEBUG
float dist = m.getClustersDistance(nearestClusters[0], nearestClusters[1]);
int c1Index = m.getClusterIndex(nearestClusters[0]),
c2Index = m.getClusterIndex(nearestClusters[1]);
// 3. merge dei cluster trovati
ICluster mergedCluster = this.mergeCluster(m, nearestClusters[0], nearestClusters[1]);
System.out.println("\n\n" + m);
System.out.println(
"\n\nMerging clusters (distance = "
+ dist
+ "):\n"
+ nearestClusters[0]
+ "\n"
+ nearestClusters[1] /* + "\nto:\n" + mergedCluster*/);
System.out.println("Indexes: " + c1Index + " " + c2Index);
dendrogram.add(mergedCluster, nearestClusters[0], nearestClusters[1]);
// 4. procede dal punto 2 per (n-1) iterazioni (altezza dendogramma pari a n = numero di
// documenti)
}
return dendrogram;
}
public static void main(String[] args) {
// Define data
DataPoint[] elements = new DataPoint[5];
elements[0] = new DataPoint("A", new double[] {});
elements[1] = new DataPoint("B", new double[] {});
elements[2] = new DataPoint("C", new double[] {});
elements[3] = new DataPoint("D", new double[] {});
elements[4] = new DataPoint("E", new double[] {});
double[][] a =
new double[][] {
{0, 1, 2, 2, 3}, {1, 0, 2, 4, 3}, {2, 2, 0, 1, 5}, {2, 4, 1, 0, 3}, {3, 3, 5, 3, 0}
};
AverageLinkAlgorithm ca = new AverageLinkAlgorithm(elements, a);
Dendrogram dnd = ca.cluster();
dnd.printAll();
}
