The library for statistical analysis.
Switch branches/tags
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
Bellona
Images/Clustering
Old
Published
Samples/ClusteringSample
Tools
.gitattributes
.gitignore
LICENSE
README.md

README.md

Bellona.Analysis

The library for statistical analysis.

Setup

To install Bellona.Analysis, run the following command in the Package Manager Console on Visual Studio:

Install-Package Bellona.Analysis

NuGet Gallery | Bellona.Analysis

Usage of Clustering

Simplest Way

The following code creates a trained clustering model with colors data.

// Colors data.
var colors = new[] { Color.Red, Color.Orange, Color.Yellow, Color.Green, Color.Blue, Color.Indigo, Color.Violet };

// Initializes and trains a clustering model.
// The lambda expression is to extract features from each color by an array of System.Double.
var model = ClusteringModel.CreateAuto<Color>(c => new double[] { c.R, c.G, c.B })
    .Train(colors);

The number of clusters doesn't need to be specified for CreateAuto method.

Remark that ClusteringModel<T> objects are immutable. CreateAuto method returns an empty model, and Train method returns a trained model. So use method chaining.

Use Result

You can access the trained result via Clusters property.

// Gets a cluster in the trained model.
var cluster0 = model.Clusters[0];
Console.WriteLine(cluster0.Id);

// Enumerates colors in cluster0.
foreach (var record in cluster0.Records)
    Console.WriteLine(record.Element.Name);

Specify Parameters

// Specifies the maximum number of clusters and the maximum standard score in σ.
var model = ClusteringModel.CreateAuto<Color>(c => new double[] { c.R, c.G, c.B })
    .Train(colors, 20, 1.5)
    .Train(colors2, 30, 1.4);

Use CreateFromNumber method to fix the number of clusters.

// Specifies the number of clusters and the maximum number of iterations.
var model = ClusteringModel.CreateFromNumber<Color>(c => new double[] { c.R, c.G, c.B }, 10)
    .Train(colors, 30);

Assign Element

Remark that Assign method doesn't train the model with the new data.

// Assigns the gold to the suitable cluster.
var cluster = model.Assign(Color.Gold);

Convert Model Type

var autoModel = ClusteringModel.CreateAuto<Color>(c => new double[] { c.R, c.G, c.B })
    .Train(colors);

// Converts AutoClusteringModel to ClusteringModel.
var fixedModel = autoModel.ToFixedModel()
    .Train(colors2);
var fixedModel = ClusteringModel.CreateFromNumber<Color>(c => new double[] { c.R, c.G, c.B }, 10)
    .Train(colors);

// Converts ClusteringModel to AutoClusteringModel.
var autoModel = fixedModel.ToAutoModel()
    .Train(colors2);

Samples

The sample source code is ClusteringSample.

Clustering colors in the System.Drawing.Color structure:
ColorClusters

Clustering the prefectural capitals in Japan by their positions:
CityClusters

Prerequisites

  • .NET Framework 4.5

Release Notes

  • v1.0.10 Add the features for clustering.
  • v1.1.13 Add minor utility methods.
  • v1.1.14 Fix bugs.