//get nearest cluster
private int GetNearestCluster(GenericVector vector)
{
var clusterid = Centroids
.OrderBy(v => GenericVector.Distance(vector, v.Value))
.Select(v => v.Key)
.FirstOrDefault();
return(clusterid);
}
public static float GetAngle(GenericVector a, GenericVector b)
{
var x = DotProduct(a, b) / (a.VectorLength() * b.VectorLength());
if (x > 1 || x < -1)
{
return(0);
}
return((float)Math.Acos(x));
}
public static float DotProduct(GenericVector vectorA, GenericVector vectorB)
{
if (vectorA.Size != vectorB.Size)
{
throw new Exception("GenericVector a size of dotProduct not equal to GenericVector b size");
}
var aTimesBpoints = vectorA.Points.Select((t, i) => t * vectorB.Points[i]).ToList();
return(aTimesBpoints.Sum());
}
public double CalculateSumofSquaredErrors()
{
var orderedclusters = Dataset.GroupBy(v => v.Cluster).OrderBy(v => v.Key).ToList();
var sse = (from cluster in orderedclusters
let clustercenter = Centroids[cluster.Key]
from point in cluster
select Math.Pow(GenericVector.Distance(clustercenter, point), 2)).Sum();
return(sse);
}
public GenericVector Sum(GenericVector vectorToSum)
{
if (Size != vectorToSum.Size)
{
throw new Exception("GenericVector size of vectorToSum not equal to instance vector size");
}
for (var i = 0; i < Size; i++)
{
Points[i] += vectorToSum.Points[i];
}
return(this);
}
//STATIC GENERICVECTOR METHODS
public static GenericVector Sum(GenericVector vectorA, GenericVector vectorB)
{
if (vectorA.Size != vectorB.Size)
{
throw new Exception("GenericVector size of vectorToSum not equal to instance vector size");
}
var summedVector = new GenericVector(vectorA.Size);
for (var i = 0; i < summedVector.Size; i++)
{
summedVector.Points[i] = vectorA.Points[i] + vectorB.Points[i];
}
return(summedVector);
}
public static void Main(string[] args)
{
var dataDictionary = new Dictionary <int, GenericVector>();
//read csv file
var lines = File.ReadAllLines("WineData.csv")
.Select(v => v.Split(',').Select(float.Parse).ToList())
.ToList();
//flip data and add to genericvector list
foreach (var line in lines)
{
for (var j = 0; j < line.Count; j++)
{
if (!dataDictionary.ContainsKey(j))
{
dataDictionary[j] = new GenericVector();
}
dataDictionary[j].Add(line[j]);
}
}
var kMeanses = new List <KMeans>();
//init kmeans ad run it.
for (var i = 0; i < 25; i++)
{
var kMeans = new KMeans
{
Iterations = 100,
Dataset = dataDictionary.Values.ToList(),
Clusters = 4
};
kMeans.Run();
kMeanses.Add(kMeans);
}
var lowestKmeans = kMeanses.Aggregate(
(minItem, nextItem) => minItem.Sse < nextItem.Sse ? minItem : nextItem);
lowestKmeans.PrintClusterInfo();
}
public static bool NotEqual(GenericVector a, GenericVector b)
{
return(a.Points.Where((value, index) => value != b.Points[index]).Any());
}
public static double Distance(GenericVector a, GenericVector b)
{
var aMinusBpoints = a.Points.Select((t, i) => t - b.Points[i]).ToList();
return(Math.Sqrt(aMinusBpoints.Sum(item => Math.Pow(item, 2))));
}
private static bool CentroidsChanged(IEnumerable <GenericVector> a, IReadOnlyList <GenericVector> b)
{
return(a.Where((item, index) => GenericVector.NotEqual(item, b[index])).Any());
}
