Search code examples
javacluster-analysiselki

Very Simple K-means clustering example in ELKI

I'm currently trying to use the KMeans Clustering functionality provided by the elki library.

This is what I came up with:

    double[][] dblArray = new double[100][10]  // 100 10-dimensional data points

    //populate array...        

    KMeansInitialization<NumberVector<Double>> kinit = new FirstKInitialMeans<>();

    KMeansLloyd<NumberVector<Double>, DoubleDistance> kmeans 
        = new KMeansLloyd<NumberVector<Double>, DoubleDistance>(EuclideanDistanceFunction.STATIC, K, KMEANSMAXITER, kinit);

    DatabaseConnection dbc = new ArrayAdapterDatabaseConnection(dblArray));
    Database d = new StaticArrayDatabase(dbc, null);

    kmeans.run(d);

Elki gives me:

de.lmu.ifi.dbs.elki.data.type.NoSupportedDataTypeException: No data type found satisfying: NumberVector,field AND NumberVector Available types: at de.lmu.ifi.dbs.elki.database.AbstractDatabase.getRelation(Unknown Source) at de.lmu.ifi.dbs.elki.algorithm.AbstractAlgorithm.run(Unknown Source)

Solution

  • Don't forget to initialize your database:

    d.initialize();

    at this point, data will be fetched from the database connections, and indexes will be built.

    If you forget to initialize your database, it will remain empty.