// this method compute distance between clusters thas has subclusters (cluster2 represents the new cluster)
public static double ComputeNodeDistance(ClusterNode node1, ClusterNode node2, DissimilarityMatrix dissimilarityMatrix, Strategy strategy)
{
ClusterNode node21 = node2.NodeAt(0);
ClusterNode node22 = node2.NodeAt(1);
double distance1 = dissimilarityMatrix.ReturnClusterPairDistance(new ClusterNodePair(node1, node21));
double distance2 = dissimilarityMatrix.ReturnClusterPairDistance(new ClusterNodePair(node1, node22));
switch (strategy)
{
case Strategy.MinLinkage:
return(distance1 < distance2 ? distance1 : distance2);
case Strategy.MaxLinkage:
return(distance1 > distance2 ? distance1 : distance2);
case Strategy.AverageLinkage:
return((distance1 + distance2) / 2);
case Strategy.AverageLinkageWeighted:
return(distance1 * node21.TotalLeafsCount / node2.TotalLeafsCount +
distance2 * node22.TotalLeafsCount / node2.TotalLeafsCount);
default:
return(0);
}
}
private void BuildHierarchicalClustering(int indexNewNode, ClusterDistance.Strategy strategy, int k)
{
ClusterNodePair closestClusterPair = dissimilarityMatrix.GetClosestClusterPair();
ClusterNode newNode = new ClusterNode();
newNode.Add(closestClusterPair.Cluster1);
newNode.Add(closestClusterPair.Cluster2);
newNode.Id = indexNewNode;
newNode.Distance = dissimilarityMatrix.ReturnClusterPairDistance(closestClusterPair);
newNode.UpdateTotalLeafs();
nodeCollection.Remove(closestClusterPair.Cluster1);
nodeCollection.Remove(closestClusterPair.Cluster2);
UpdateDissimilarityMatrix(newNode, strategy);
nodeCollection.Add(newNode);
if (nodeCollection.Count > k)
{
BuildHierarchicalClustering(indexNewNode + 1, strategy, k);
}
}
