Mean Shift

Mean Shift is a clustering algorithm that does not require specifying the number of clusters in advance. It works by iteratively shifting each data point towards the region with the highest density of surrounding points, calculated using a "window" of a defined radius. In the end, points that converge to the same dense region form a cluster. This method is useful for automatically detecting the number of clusters and works well with data of varying shapes, but it can be computationally more expensive on large datasets.

EXAMPLE OF USE

uses
  UMeanShift;
  
procedure ExampleClustering;
begin
  ShowArray(MeanShift('C:\Delphai\Delphai\Datasets\Iris-Clustering.csv', 0.2, 0.1));
end;

CLASSES AND METHODS GUIDE

• function MeanShift(aData : String; aBandwidth, aEpsilon : Double; aMaxIterations: Integer = 300; aHasHeader : Boolean = True): TArray; or

• function MeanShift(aData : TDataSet; aBandwidth, aEpsilon : Double; aMaxIterations: Integer = 300): TArray; : defines the groups.

  • aData : CSV file path or TDataSet object that contains the data that will be used for training.

  • aBandwidth (Largura de Banda) : determines the radius used to define the neighborhood of each point during clustering.

  • aEpsilon (Tolerância de Convergência) : specifies the minimum difference threshold to consider that the centroids have converged.

  • aMaxIterations (Número Máximo de Iterações) : defines the maximum number of cycles allowed to adjust the clusters and calculate the centroids.

  • aHasHeader : indicates whether the file has a header.