private Algorithms.GravitationalClustering PredictIrisDataSimProto()
{
var data = File.ReadAllLines("iris.csv")
.Skip(1)
.Select(x =>
{
var left = x.Split(',').Skip(1).Select(y => double.Parse(y)).ToArray();
var right = int.Parse(x.Split(',').Last());
return(Tuple.Create(left, right));
})
.ToArray();
Algorithms.GravitationalClustering clustering = new Algorithms.GravitationalClustering(0.01, 0.1, 100, new SquaredEuclideanDistance());
return(RunProtoGCAlgoritym(ref data, clustering, true));
}
private Algorithms.GravitationalClustering PredictOllivetiDataSimProto()
{
int index = 0;
var pred = File.ReadAllLines("olevitti_p.csv");
var data = File.ReadAllLines("olevitti.csv")
.Select(x =>
{
var left = x.Split(',').Skip(1).Select(y => double.Parse(y)).ToArray();
return(Tuple.Create(left, (int)double.Parse(pred[index++])));
})
.ToArray();
Algorithms.GravitationalClustering clustering = new Algorithms.GravitationalClustering(10, 0.1, 1, new SquaredEuclideanDistance());
return(RunProtoGCAlgoritym(ref data, clustering, true));
}
private Algorithms.GravitationalClustering PredictDigitsDataProb()
{
var data = File.ReadAllLines("digits.csv")
.Skip(1)
.Select(x =>
{
var left = x.Split(',').Skip(1).Select(y => double.Parse(y)).ToArray();
var right = int.Parse(x.Split(',').Last());
return(Tuple.Create(left, right));
})
.ToArray();
Algorithms.GravitationalClustering clustering = new Algorithms.GravitationalClustering(40, 0.01, 100, new SquaredEuclideanDistance());
return(RunGCAlgorithm(ref data, clustering, false, 50));
}
public void DrawClusters(Algorithms.GravitationalClustering clust)
{
for (int i = 0; i < 4; i++)
{
for (int x = i; x < 4; x++)
{
if (i != x)
{
Graphics g = this.pictureBox1.CreateGraphics();
foreach (var item in clust.EnumeratePlanets())
{
this.DrawCircle(g, Pens.Red, new Point(
(int)(item.Position[i] * this.ScaleCluster + this.ShiftCluster),
(int)(item.Position[x] * this.ScaleCluster + 300)),
(int)((item.Radius * ScaleRadius) * this.ScaleCluster * this.ScaleCluster));
}
this.ShiftCluster += 100;
}
}
}
}
private Algorithms.GravitationalClustering PredictBreastCancerSimProto()
{
double[][] input = new double[569][];
for (int x = 0; x < input.Length; x++)
{
input[x] = new double[32];
}
double[] output = new double[569];
var data = File.ReadAllLines("wdbc.txt")
.Select(x =>
{
var left = x.Split(',').Skip(2).Select(y => double.Parse(y)).ToArray();
var right = x.Split(',').Skip(1).First().First() == 'M' ? 0 : 1;
return(Tuple.Create(left, right));
})
.ToArray();
Algorithms.GravitationalClustering clustering = new Algorithms.GravitationalClustering(5000, 0.01, 100, new SquaredEuclideanDistance());
return(RunProtoGCAlgoritym(ref data, clustering, true));
}
private Algorithms.GravitationalClustering RunProtoGCAlgoritym(ref Tuple <double[], int>[] data, Algorithms.GravitationalClustering clustering, bool sim)
{
var newData = this.GetProto(data);
foreach (var item in newData)
{
clustering.OnlineTrainSingular(item.Item1, item.Item2);
}
Debug.WriteLine("Number of Clusters " + clustering.EnumeratePlanets().Count());
var d = data.Select(xx =>
{
int outP = 0;
if (sim)
{
outP = clustering.PredictSimulation(xx.Item1);
}
else
{
outP = clustering.PredictProbabilistic(xx.Item1);
}
Debug.WriteLine("Predicted Class " + outP);
if (outP == xx.Item2)
{
return(1);
}
return(0);
}).Sum();
#if DEBUG
{
Debug.WriteLine("Percent Accuracy " + ((double)d / ((double)data.Length)).ToString());
}
#else
{
MessageBox.Show(((double)d / ((double)data.Length)).ToString());
}
#endif
return(clustering);
}
private Algorithms.GravitationalClustering RunGCAlgorithm(ref Tuple <double[], int>[] data, Algorithms.GravitationalClustering clustering, bool sim, int rand = 4)
{
RandomizeArray(ref data, rand);
int take = data.Length / 2;
foreach (var item in data.Take(take))
{
clustering.OnlineTrainSingular(item.Item1, item.Item2);
}
Debug.WriteLine("Number of Clusters " + clustering.EnumeratePlanets().Count());
var d = data.Skip(take).Select(xx =>
{
int outP = 0;
if (sim)
{
outP = clustering.PredictSimulation(xx.Item1);
}
else
{
outP = clustering.PredictProbabilistic(xx.Item1);
}
Debug.WriteLine("Predicted Class " + outP);
if (outP == xx.Item2)
{
return(1);
}
return(0);
}).Sum();
#if DEBUG
{
Debug.WriteLine("Percent Accuracy " + ((double)d / ((double)data.Length - take)).ToString());
}
#else
{
MessageBox.Show(((double)d / ((double)data.Length - take)).ToString());
}
#endif
return(clustering);
}
