Package org.zkoss.chart

Class ClusterLayoutAlgorithm

  • All Implemented Interfaces:
    Serializable, OptionDataListener, org.zkoss.json.JSONAware
    public class ClusterLayoutAlgorithm
extends Optionable
    Options for layout algorithm. Inside there are options to change the type of the algorithm, gridSize, distance or iterations.

    All the options in this class support DynamicalAttribute.

    Since:
    10.2.1.0
    Author:
    jumperchen
    See Also:
    Serialized Form

    • Constructor Detail

      • ClusterLayoutAlgorithm

        public ClusterLayoutAlgorithm()

    • Method Detail

      • getDistance

        public Object getDistance()
        When type is set to kmeans, distance is a maximum distance between point and cluster center so that this point will be inside the cluster. The distance is either a number defining pixels or a percentage defining a percentage of the plot area width.

        Default: 40

      • setDistance

        public void setDistance​(String distance)
        When type is set to kmeans, distance is a maximum distance between point and cluster center so that this point will be inside the cluster. The distance is either a number defining pixels or a percentage defining a percentage of the plot area width.

        Default: 40

        See Also:
        setDistance(Number)

      • setDistance

        public void setDistance​(Number distance)
        When type is set to kmeans, distance is a maximum distance between point and cluster center so that this point will be inside the cluster. The distance is either a number defining pixels or a percentage defining a percentage of the plot area width.

        Default: 40

        See Also:
        setDistance(String)

      • getGridSize

        public Object getGridSize()
        When type is set to the grid, gridSize is a size of a grid square element either as a number defining pixels, or a percentage defining a percentage of the plot area width.

        Default: 50

      • setGridSize

        public void setGridSize​(String gridSize)
        When type is set to the grid, gridSize is a size of a grid square element either as a number defining pixels, or a percentage defining a percentage of the plot area width.

        Default: 50

        See Also:
        setGridSize(Number)

      • setGridSize

        public void setGridSize​(Number gridSize)
        When type is set to the grid, gridSize is a size of a grid square element either as a number defining pixels, or a percentage defining a percentage of the plot area width.

        Default: 50

        See Also:
        setGridSize(String)

      • getIterations

        public Number getIterations()
        When type is set to kmeans, iterations are the number of iterations that this algorithm will be repeated to find clusters positions.

        Default: null

      • setIterations

        public void setIterations​(Number iterations)
        When type is set to kmeans, iterations are the number of iterations that this algorithm will be repeated to find clusters positions.

        Default: null

      • getKmeansThreshold

        public Number getKmeansThreshold()
        When type is set to null and there are more visible points than the kmeansThreshold the grid algorithm is used to find clusters, otherwise kmeans. It ensures good performance on large datasets and better clusters arrangement after the zoom.

        Default: 100

      • setKmeansThreshold

        public void setKmeansThreshold​(Number kmeansThreshold)
        When type is set to null and there are more visible points than the kmeansThreshold the grid algorithm is used to find clusters, otherwise kmeans. It ensures good performance on large datasets and better clusters arrangement after the zoom.

        Default: 100

      • getType

        public Object getType()
        Returns the type of the algorithm used to combine points into a cluster.

        Default: null.

      • setType

        public void setType​(String type)
        Sets the type of the algorithm used to combine points into a cluster.

        There are three available algorithms:

        • grid - grid-based clustering technique. Points are assigned to squares of set size depending on their position on the plot area. Points inside the grid square are combined into a cluster. The grid size can be controlled by gridSize property (grid size changes at certain zoom levels).
        • kmeans - based on K-Means clustering technique. In the first step, points are divided using the grid method (distance property is a grid size) to find the initial amount of clusters. Next, each point is classified by computing the distance between each cluster center and that point. When the closest cluster distance is lower than distance property set by a user the point is added to this cluster otherwise is classified as noise. The algorithm is repeated until each cluster center not change its previous position more than one pixel. This technique is more accurate but also more time consuming than the grid algorithm, especially for big datasets.
        • optimizedKmeans - based on K-Means clustering technique. This algorithm uses k-means algorithm only on the chart initialization or when chart extremes have greater range than on initialization. When a chart is redrawn the algorithm checks only clustered points distance from the cluster center and rebuild it when the point is spaced enough to be outside the cluster. It provides performance improvement and more stable clusters position yet can be used rather on small and sparse datasets.

        By default, the algorithm depends on visible quantity of points and kmeansThreshold. When there are more visible points than the kmeansThreshold the grid algorithm is used, otherwise kmeans.

        Parameters:
        type - either grid, kmeans, and optimizedKmeans

      • setType

        public void setType​(org.zkoss.json.JavaScriptValue type)
        Sets the custom clustering algorithm can be added by assigning a callback function as the type property. This function takes an array of processedXData, processedYData, processedXData indexes and layoutAlgorithm options as arguments and should return an object with grouped data.

        The algorithm should return an object like that: 

        
{
  clusterId1: [{
      x: 573,
      y: 285,
      index: 1 // point index in the data array
  }, {
      x: 521,
      y: 197,
      index: 2
  }],
  clusterId2: [{
      ...
  }]
  ...
 }
 }

        clusterId (example above - unique id of a cluster or noise) is an array of points belonging to a cluster. If the array has only one point or fewer points than set in cluster.minimumClusterSize it won't be combined into a cluster.

        See Also:
        setType(String)