public void Assign_Clusters()
{
var o = new ClusterAlgorithmOptions(2);
ClusterData cd = new ClusterData(new double[3][] { new double[2], new double[2], new double[2] }, new string[2]);
ClusterResult cr = new ClusterResult();
cr.ClusterAssignment = new int[3] {
0, 1, 0
};
cr.ClusterMeanValues = new double[2][];
cr.ClusterMeanValues[0] = new double[2] {
100, 1000
};
cr.ClusterMeanValues[1] = new double[2] {
5, 10
};
cr.NormalizedData = new double[3][];
cr.NormalizedData[0] = new double[2] {
80, 900
};
cr.NormalizedData[1] = new double[2] {
200, 2000
};
cr.NormalizedData[2] = new double[2] {
7, 20
};
KmeansClusterAlgorithm.UpdateClusterAssignment(cr, cd, o);
Assert.AreEqual(cr.ClusterAssignment[0], 0);
Assert.AreEqual(cr.ClusterAssignment[1], 0);
Assert.AreEqual(cr.ClusterAssignment[2], 1);
}
public void Initialize_Clusters()
{
var x = new double[10][];
for (int i = 0; i < x.Length; i++)
{
x[i] = new double[5];
}
ClusterData d = new ClusterData(x, new string[] { "a1", "a2", "a3", "a4", "a5" });
ClusterAlgorithmOptions options = new ClusterAlgorithmOptions(3, 123, true, 10);
//should provide initialized mean values and assignment
ClusterResult r = KmeansClusterAlgorithm.InitializeClusters(d, options);
int[] count = new int[3];
for (int i = 0; i < d.RawData.Length; i++)
{
int idx = r.ClusterAssignment[i];
count[idx] += 1;
}
for (int i = 0; i < options.NumberOfClusters; i++)
{
//check if every cluster received at least one data point
Assert.IsTrue(count[i] > 0);
//check if the column mean values are initialized
Assert.IsTrue(r.ClusterMeanValues[i].Length == d.RawData[0].Length);
}
}
public void Can_Find_Minimum()
{
double[] data = new double[] { -5, 10, 3, 100, -6.01, -6.02, 7, -3 };
int expectedIndex = 5;
int foundIndex = KmeansClusterAlgorithm.FindMinimumIndex(data);
Assert.AreEqual(expectedIndex, foundIndex);
}
public void TestClusterAlgorithmOnExample()
{
ClusterData data = new ClusterData(GetTestData(), new string[] { "X", "Y" });
ClusterAlgorithmOptions options = new ClusterAlgorithmOptions(5);
var result = KmeansClusterAlgorithm.Analyze(data, options);
Assert.IsTrue(result.ClusterAssignment.Length > 0);
Trace.WriteLine($"Iterations: {result.Iterations}");
Trace.WriteLine(result.PrintCsvResult());
}
public void Distance_Calculation_Euclidean()
{
double[] means = new double[] { 3, 4, 5 };
double[] data = new double[] { 10, 10, 10 };
double[] weights = new double[] { 1, 1, 1 };
double result = 0;
for (int i = 0; i < means.Length; i++)
{
result = result + weights[i] * Math.Pow(means[i] - data[i], 2);
}
result = Math.Sqrt(result);
Assert.AreEqual(result, KmeansClusterAlgorithm.GetDistance(means, data, weights));
}
public void TestBigSet()
{
int cols = 5;
int rows = 5000;
string[] attributes = new string[cols];
for (int i = 0; i < cols; i++)
{
attributes[i] = "Col_" + i.ToString();
}
ClusterAlgorithmOptions options = new ClusterAlgorithmOptions(3, true);
ClusterData data = new ClusterData(GetRandomDataSet(rows, cols, 2000), attributes);
ClusterResult result = KmeansClusterAlgorithm.Analyze(data, options);
Assert.IsTrue(result.Iterations > 0);
Trace.WriteLine($"Iterations: {result.Iterations}");
Trace.WriteLine(result.PrintCsvResult());
}
public void Calculate_Means()
{
double[][] data = new double[5][];
//1. cluster
data[0] = new double[] { 2, 7 };
data[1] = new double[] { 3, 3 };
//2. cluster
data[2] = new double[] { 1, 100 };
data[3] = new double[] { 2, 200 };
data[4] = new double[] { 3, 300 };
ClusterData cd = new ClusterData(data, new string[] { "a1", "a2" });
ClusterAlgorithmOptions options = new ClusterAlgorithmOptions(2, false);
ClusterResult r = new ClusterResult();
r.ClusterAssignment = new int[] { 0, 0, 1, 1, 1 };
r.NormalizedData = data;
r.ClusterMeanValues = new double[2][];
r.ClusterMeanValues[0] = new double[2] {
0, 0
};
r.ClusterMeanValues[1] = new double[2] {
0, 0
};
KmeansClusterAlgorithm.UpdateClusterMeanValues(r, cd, options);
double exp_First = 2.5;
double exp_Second = 5.0;
Assert.AreEqual(exp_First, r.ClusterMeanValues[0][0]);
Assert.AreEqual(exp_Second, r.ClusterMeanValues[0][1]);
exp_First = 2d;
exp_Second = 200d;
Assert.AreEqual(exp_First, r.ClusterMeanValues[1][0]);
Assert.AreEqual(exp_Second, r.ClusterMeanValues[1][1]);
}
private async void ExportButton_Click(object sender, RoutedEventArgs e)
{
ClusterButton.IsEnabled = false;
ClusterButton.IsEnabled = false;
LoadCsvButton.IsEnabled = false;
BusyIndication.IsActive = true;
int clusters = (int)(ClusterCountRange.UpperValue);
if (MessageBox.Show($"Run with {clusters} Clusters?", "Get Results", MessageBoxButton.YesNo, MessageBoxImage.Question)
== MessageBoxResult.Yes)
{
int runs = (int)(RandomRuns.Value.Value);
bool normalize = NormalizeSwitch.IsChecked.Value;
var analysis = await Task.Run(() => ClusterSolver.Cluster(clusterData, clusters, clusters, runs, normalize));
var seed = analysis[0].BestSeed;
var options = new ClusterAlgorithmOptions(clusters, seed, normalize);
var result = await Task.Run(() => KmeansClusterAlgorithm.Analyze(clusterData, options));
SaveFileDialog saveFileDialog = new SaveFileDialog();
saveFileDialog.Filter = "Csv files (*.csv)|*.csv|Text files (*.txt)|*.txt";
if (saveFileDialog.ShowDialog() == true)
{
using (var writer = new StreamWriter(File.Create(saveFileDialog.FileName)))
{
string csv = result.PrintCsvResult();
writer.Write(csv);
}
}
}
ClusterButton.IsEnabled = true;
ClusterButton.IsEnabled = true;
LoadCsvButton.IsEnabled = true;
BusyIndication.IsActive = false;
}
