public DBSCAN(int MinPts, double radius)
{
_MinPts = MinPts;
_radius = radius;
GetDistance = (a, b) => DistanceFunctions.HaversineDistance(a, b).DistanceInMeters;
IsCloseEnough = (a, b, c) => (GetDistance(a, b) < (Math.Exp(-Math.Pow(c, 0.6) / 40) * _radius));
}
public List <Cluster <T> > Cluster(List <Coord> points, List <T> data)
{
int n = points.Count();
// we need n x c - many random numbers to initialize membership
Random random = new Random(7);
double[,] U = new double[n, _c];
for (int i = 0; i < n; i++)
{
for (int j = 0; j < _c; j++)
{
U[i, j] = random.NextDouble();
}
}
List <Coord> centers = new List <Coord>();
double lowerLat = points.Min(x => x.Latitude);
double lowerLon = points.Min(x => x.Longitude);
double upperLat = points.Max(x => x.Latitude);
double upperLon = points.Max(x => x.Longitude);
for (int i = 0; i < _c; i++)
{
double RandomLat = random.NextDouble() * (upperLat - lowerLat) + lowerLat;
double RandomLon = random.NextDouble() * (upperLon - lowerLon) + lowerLon;
Coord randomCenter = new Coord(RandomLat, RandomLon);
centers.Add(randomCenter);
}
for (int p = 0; p < _maxiters; p++)
{
//update centers
for (int i = 0; i < _c; i++)
{
double firstvalue_lat = SumColumnExponentiatedAndMultiplyWithData_Latitude(U, points, i, _m);
double firstvalue_lon = SumColumnExponentiatedAndMultiplyWithData_Longitude(U, points, i, _m);
double secondvalue = SumColumnExponentiated(U, i, _m); //sum over the n rows of U raised to power m
if (secondvalue < 0.00001)
{
secondvalue = 0.00001;
}
//putting this in to see what happens
double new_lat = firstvalue_lat / secondvalue;
double new_lon = firstvalue_lon / secondvalue;
if (!Double.IsNaN(new_lat) && !Double.IsNaN(new_lat))
{
centers[i].Latitude = new_lat;
centers[i].Longitude = new_lon;
}
else
{
//Console.WriteLine("Paused");
}
}
//update membership matrix U
for (int i = 0; i < n; i++)
{
for (int j = 0; j < _c; j++)
{
//double theijconstantweneed = Math.Pow(DistanceFunctions.HaversineDistance(points[i], centers[j]).DistanceInMeters, 2 / (_m - 1));
double theijconstantweneed = Math.Pow(DistanceFunctions.HaversineDistance(points[i], centers[j]).DistanceInMeters, 2 / (_m - 1));
//double theijconstantweneed = Math.Pow(DistanceFunctions.LawOfCosinesDistance(points[i], centers[j]).DistanceInMeters, 2 / (_m - 1));
//double theijconstantweneed = Math.Pow(DistanceFunctions.ApproximateDistance(points[i], centers[j]).DistanceInMeters, 2 / (_m - 1));
double denominator = 0.0;
for (int k = 0; k < _c; k++)
{
//denominator += theijconstantweneed / Math.Pow(DistanceFunctions.HaversineDistance(points[i], centers[k]).DistanceInMeters, 2 / (_m - 1));
denominator += theijconstantweneed / Math.Pow(DistanceFunctions.HaversineDistance(points[i], centers[k]).DistanceInMeters, 2 / (_m - 1));
//denominator += theijconstantweneed / Math.Pow(DistanceFunctions.LawOfCosinesDistance(points[i], centers[k]).DistanceInMeters, 2 / (_m - 1));
//denominator += theijconstantweneed / Math.Pow(DistanceFunctions.ApproximateDistance(points[i], centers[k]).DistanceInMeters, 2 / (_m - 1));
}
//if (denominator == 0)
//{
// denominator = 0.00001;
//}
U[i, j] = 1 / denominator;
}
}
}
var asdsjhf = new List <T> [_c];
for (int i = 0; i < _c; i++)
{
asdsjhf[i] = new List <T>();
}
for (int i = 0; i < n; i++)
{
int k = 0;
var not_smallest = U[i, k];
for (int j = 0; j < _c; j++)
{
if (U[i, j] > not_smallest)
{
not_smallest = U[i, j];
k = j;
}
}
asdsjhf[k].Add(data[i]);
}
var retstuff = new List <Cluster <T> >();
for (int j = 0; j < _c; j++)
{
var gjhh = new Cluster <T>(centers[j], asdsjhf[j]);
retstuff.Add(gjhh);
}
return(retstuff);
}