// Calcula a menor distancia entre todos os grupos
private static double MinDistanceBetweenClusters(Cluster[] clusters)
{
DissimilarityMatrix dissimilarityMatrix = new DissimilarityMatrix();
for (int i = 0; i < clusters.Count() - 1; i++)
{
for (int j = i + 1; j < clusters.Count(); j++)
{
CalculateDistanceBetweenClusters(clusters[i], clusters[j], dissimilarityMatrix);
}
}
return(dissimilarityMatrix.GetLowestDistance());
}
/// <summary>
/// Computes the distance.
/// </summary>
/// <param name="cluster1">The cluster1.</param>
/// <param name="cluster2">The cluster2.</param>
/// <param name="dissimilarityMatrix">The dissimilarity matrix.</param>
/// <param name="strategy">The strategy.</param>
/// <returns>Distance between clusters</returns>
public static double ComputeDistance(Cluster cluster1, Cluster cluster2, DissimilarityMatrix dissimilarityMatrix, MergeStrategy strategy)
{
double distance = 0;
var distance1 = dissimilarityMatrix.ReturnClusterPairDistance(new ClusterPair(cluster1, cluster2.GetSubCluster(0)));
var distance2 = dissimilarityMatrix.ReturnClusterPairDistance(new ClusterPair(cluster1, cluster2.GetSubCluster(1)));
// computes distance by using merge strategy
switch (strategy)
{
case MergeStrategy.SingleLinkage:
distance = _MinValue(distance1, distance2); // Min(x, y)
break;
case MergeStrategy.CompleteLinkage:
distance = _MaxValue(distance1, distance2); // Max(x, y)
break;
case MergeStrategy.AverageLinkageWpgma:
distance = (distance1 + distance2) / 2; // Avg(x, y)
break;
case MergeStrategy.AverageLinkageUpgma:
distance = ((cluster2.GetSubCluster(0).QuantityOfDataPoints * distance1) / cluster2.QuantityOfDataPoints)
+ ((cluster2.GetSubCluster(1).QuantityOfDataPoints * distance2) / cluster2.QuantityOfDataPoints); // WeightedAvg(x, y)
break;
case MergeStrategy.CentroidMethod:
cluster1.SetCentroid();
cluster2.SetCentroid();
distance = Distance.GetDistance(cluster1.Centroid, cluster2.Centroid,
DistanceMetric.SquareEuclidianDistance); // Distance of centroids
break;
case MergeStrategy.WardsMethod:
Cluster newCluster = new Cluster();
newCluster.AddSubCluster(cluster1);
newCluster.AddSubCluster(cluster2);
newCluster.SetCentroid();
distance = newCluster.GetSumOfSquaredError(DistanceMetric.EuclidianDistance)
- cluster1.GetSumOfSquaredError(DistanceMetric.EuclidianDistance)
- cluster2.GetSumOfSquaredError(DistanceMetric.EuclidianDistance);
// SEO(xy) - SEO(x) - SEO(y)
break;
}
return distance;
}
// Calcula a menor distancia entre dois grupos
private static void CalculateDistanceBetweenClusters(Cluster cluster0, Cluster cluster1, DissimilarityMatrix dissimilarityMatrix)
{
double distanceBetweenClusters = 0.0; // distancia entre dois grupos
double minDistance = Double.MaxValue; // menor distancia entre dois grupos
foreach (Pattern patternCluster0 in cluster0.GetPatterns())
{
foreach (Pattern patternCluster1 in cluster1.GetPatterns())
{
distanceBetweenClusters = Metrics.Distance.CalculatePatternDistance(patternCluster0, patternCluster1);
if (distanceBetweenClusters < minDistance)
{
minDistance = distanceBetweenClusters;
}
}
}
dissimilarityMatrix.AddClusterPairAndDistance(new ClusterPair(cluster0, cluster1), minDistance);
}
