Add CalinskiHarabazIndex

src/main/java/fr/irit/smac/amas4dc/cluster/CalinskiHarabazIndex.java

+package fr.irit.smac.amas4dc.cluster;
+
+import fr.irit.smac.amas4dc.amas.MASSettings;
+
+import java.util.List;
+
+public class CalinskiHarabazIndex<T extends DataPoint> {
+
+	public double compute(MASSettings<T> masSettings, List<ExtendedCluster<T>> clusters) {
+		var ssb = computeBetweenClusterVariance(clusters, masSettings);
+		var ssw = computeWithinClusterVariance(clusters, masSettings);
+
+		if (ssw == 0) {
+			// Avoid division by zero
+			return Double.POSITIVE_INFINITY;
+		}
+
+		return ssb / ssw;
+	}
+
+	private double computeBetweenClusterVariance(List<ExtendedCluster<T>> clusters, MASSettings<T> masSettings) {
+		var centroid = computeGlobalCentroid(clusters, masSettings);
+		var ssb = 0.0;
+
+		for (var cluster : clusters) {
+			var centroidDist = 1.0/masSettings.similarityScoreMethod().apply(cluster.getRepresentative(), centroid);
+			ssb += centroidDist * centroidDist * cluster.getContent().size();
+		}
+
+		return ssb;
+	}
+
+	private double computeWithinClusterVariance(List<ExtendedCluster<T>> clusters, MASSettings<T> masSettings) {
+		var ssw = 0.0;
+
+		for (var cluster : clusters) {
+			var clusterCentroid = cluster.getRepresentative();
+			var clusterSumDist = 0.0;
+
+			for (var point : cluster.getContent()) {
+				var dist = 1.0/masSettings.similarityScoreMethod().apply(point, clusterCentroid);
+				clusterSumDist += dist * dist;
+			}
+
+			ssw += clusterSumDist;
+		}
+
+		return ssw;
+	}
+
+	private T computeGlobalCentroid(List<ExtendedCluster<T>> clusters, MASSettings<T> masSettings) {
+		var fused = clusters.get(0).getRepresentative();
+		for (int i = 1; i < clusters.size(); i++) {
+			fused = masSettings.dataPointFuser().apply(fused, clusters.get(i).getRepresentative());
+		}
+		return fused;
+	}
+}
+