private const uint NUM_CLUSTERS = 3; // for seeds and iris data-sets
#endregion
#region Private & Protected Methods
private static void EvaluateClustering(
ISet <DataPoint> dataPoints, ILinkageCriterion <DataPoint> linkage, string linkageName, uint numClusters)
{
var clusteringAlg = new AgglomerativeClusteringAlgorithm <DataPoint>(linkage);
var clustering = clusteringAlg.GetClustering(dataPoints);
// gets cluster set according to predefined number of clusters
var clusterSet = clustering.First(cs => cs.Count == numClusters);
// gets classes for each data-point (first character of the ID in the dataset)
var pointClasses = dataPoints.ToDictionary(dataPoint => dataPoint, dataPoint => dataPoint.ID[0]);
Console.WriteLine("=============================================");
Console.WriteLine($"Evaluating {linkageName} clustering using Euclidean distance...");
// evaluates the clustering according to different criteria
var evaluations =
new Dictionary <string, double>
{
{ "Purity", new Purity <DataPoint, char>().Evaluate(clusterSet, pointClasses) },
{ "NMI", new NormalizedMutualInformation <DataPoint, char>().Evaluate(clusterSet, pointClasses) },
{ "Accuracy", new RandIndex <DataPoint, char>().Evaluate(clusterSet, pointClasses) },
{ "Precision", new Precision <DataPoint, char>().Evaluate(clusterSet, pointClasses) },
{ "Recall", new Recall <DataPoint, char>().Evaluate(clusterSet, pointClasses) },
{ "F1Measure", new FMeasure <DataPoint, char>(1).Evaluate(clusterSet, pointClasses) },
{ "F2Measure", new FMeasure <DataPoint, char>(2).Evaluate(clusterSet, pointClasses) },
{ "F05Measure", new FMeasure <DataPoint, char>(0.5).Evaluate(clusterSet, pointClasses) }
};
foreach (var evaluation in evaluations)
{
Console.WriteLine($" - {evaluation.Key}: {evaluation.Value:0.000}");
}
}
private static void PrintClusters(ISet <DataPoint> instances, ILinkageCriterion <DataPoint> linkage, string name)
{
var perfMeasure = new PerformanceMeasure();
perfMeasure.Start();
var clusteringAlg = new AgglomerativeClusteringAlgorithm <DataPoint>(linkage);
var clustering = clusteringAlg.GetClustering(instances);
perfMeasure.Stop();
Console.WriteLine("_____________________________________________");
Console.WriteLine(name);
Console.WriteLine(perfMeasure);
foreach (var clusterSet in clustering)
{
Console.WriteLine($"Clusters at distance: {clusterSet.Dissimilarity:0.00} ({clusterSet.Count})");
foreach (var cluster in clusterSet)
{
Console.WriteLine($" - {cluster}");
}
}
clustering.SaveD3DendrogramFile(Path.GetFullPath($"{name}.json"), formatting: Formatting.Indented);
}
/// <summary>
/// Creates a new instance of <see cref="AgglomerativeClusteringAlgorithm{TInstance}" /> with the given set of
/// instances and linkage
/// criterion.
/// </summary>
/// <param name="linkageCriterion">The criterion used to measure dissimilarities within and between clusters.</param>
public AgglomerativeClusteringAlgorithm(ILinkageCriterion <TInstance> linkageCriterion)
{
this.LinkageCriterion = linkageCriterion;
}
