/// <summary>
/// Builds the hierarchical clustering.
/// </summary>
/// <param name="indexNewCluster">The index new cluster.</param>
/// <param name="k">The k.</param>
/// <param name="isWithIndex">if set to <c>true</c> [is with index].</param>
private void BuildHierarchicalClustering(int indexNewCluster, int k, bool isWithIndex = false)
{
ClusterPair closestClusterPair = _dissimilarityMatrix.GetClosestClusterPair(); // gets the clusterpair with minimal distance
// creates new cluster by merging clusters from closestClusterPair
Cluster newCluster = new Cluster();
newCluster.AddSubCluster(closestClusterPair.Cluster1);
newCluster.AddSubCluster(closestClusterPair.Cluster2);
newCluster.Id = indexNewCluster;
newCluster.SetCentroid();
// removes cluster pair from _clusters
_clusters.RemoveClusterPair(closestClusterPair);
_UpdateDissimilarityMatrix(newCluster);
// add new cluster to _clusters
_clusters.AddCluster(newCluster);
if (isWithIndex) // checks is executed for calculating CH index
{
_chValue.Add(GetCHIndex()); // adds index to array of CH values
_chIndex.Add(_clusters.ClustersList.Count); // adds number of clusters for current CH value
}
// exit point of algorithm (Where _clusters count is equal to k)
if (_clusters.Count > k)
{
BuildHierarchicalClustering(indexNewCluster + 1, k, isWithIndex);
}
}
/// <summary>
/// Returns the cluster pair distance.
/// </summary>
/// <param name="clusterPair">The cluster pair.</param>
/// <returns></returns>
public double ReturnClusterPairDistance(ClusterPair clusterPair)
{
double clusterPairDistance;
clusterPairDistance = _distanceMatrix.ContainsKey(clusterPair) ? _distanceMatrix[clusterPair] : _distanceMatrix[new ClusterPair(clusterPair.Cluster2, clusterPair.Cluster1)];
return(clusterPairDistance);
}
/// <summary>
/// Builds the dissimilarity matrix.
/// </summary>
private void BuildDissimilarityMatrix()
{
_dissimilarityMatrix = new DissimilarityMatrix();
for (int i = 0; i < _clusters.Count - 1; i++)
{
for (int j = i + 1; j < _clusters.Count; j++)
{
var clusterPair = new ClusterPair(_clusters.GetCluster(i), _clusters.GetCluster(j));
var distanceBetweenTwoClusters = ClusterDistance.ComputeDistance(clusterPair.Cluster1, clusterPair.Cluster2, _distanceMetric);
_dissimilarityMatrix.AddClusterPairAndDistance(clusterPair, distanceBetweenTwoClusters); // adds distance to matrix
}
}
}
/// <summary>
/// Gets the closest cluster pair.
/// </summary>
/// <returns></returns>
public ClusterPair GetClosestClusterPair()
{
double minDistance = double.MaxValue;
ClusterPair closestClusterPair = null;
foreach (var item in _distanceMatrix)
{
if (item.Value < minDistance)
{
minDistance = item.Value;
closestClusterPair = item.Key;
}
}
return(closestClusterPair);
}
/// <summary>
/// Removes the cluster pair.
/// </summary>
/// <param name="clusterPair">The cluster pair.</param>
public void RemoveClusterPair(ClusterPair clusterPair)
{
_distanceMatrix.Remove(_distanceMatrix.ContainsKey(clusterPair)
? clusterPair
: new ClusterPair(clusterPair.Cluster2, clusterPair.Cluster1));
}
/// <summary>
/// Adds the cluster pair and distance.
/// </summary>
/// <param name="clusterPair">The cluster pair.</param>
/// <param name="distance">The distance.</param>
public void AddClusterPairAndDistance(ClusterPair clusterPair, double distance)
{
_distanceMatrix.Add(clusterPair, distance);
}
/// <summary>
/// Removes the cluster pair.
/// </summary>
/// <param name="clusterPair">The cluster pair.</param>
public void RemoveClusterPair(ClusterPair clusterPair)
{
RemoveCluster(clusterPair.Cluster1);
RemoveCluster(clusterPair.Cluster2);
}
