public static DecisionTree CreateTree(int[] indexes, double[,] xy, int nclasses, double pcoeff, double vcoeff)
{
if (xy == null)
{
throw new ArgumentException("xy is null", nameof(xy));
}
int npoints = xy.GetLength(0);
int nvars = xy.GetLength(1) - 1;
int modNpoints = (int)(npoints * pcoeff); // столько значений надо нагенерировать
int modNvars = (int)(nvars * vcoeff); // столько переменных используем
if (modNvars < 1)
{
throw new ArgumentException("vcoeff too small", nameof(vcoeff));
}
int[] vidxes = Enumerable.Range(0, nvars).ToArray();
if (modNvars < nvars)
{
vidxes = Enumerable.Range(0, nvars).OrderBy(c => RandomGen.GetDouble()).Take(modNvars).ToArray();
}
double[,] nxy = new double[modNpoints, modNvars + 1]; // сами значения
int nk = 0; // столько нагенерировали
var exists = new Dictionary <int, int>();
if (indexes == null)
{
// basic tree
while (nk < modNpoints)
{
for (int i = 0; i < modNpoints; i++)
{
int sn = (int)(RandomGen.GetDouble() * npoints); // selection distribution
if (sn >= npoints)
{
continue;
}
if (exists.ContainsKey(sn))
{
continue; // такой ключ уже был
}
exists.Add(sn, 0);
for (int j = 0; j < modNvars; j++)
{
nxy[i, j] = xy[sn, vidxes[j]];
}
nxy[i, modNvars] = xy[sn, nvars];
nk++;
if (nk >= modNpoints)
{
break;
}
}
if (nk >= modNpoints)
{
break;
}
}
}
else
{
// tree, with distribution selection
while (nk < modNpoints)
{
for (int i = 0; i < modNpoints; i++)
{
int sn = (int)(RandomGen.GetTrangle() * npoints); // selection distribution
if (sn >= indexes.Length)
{
continue;
}
if (exists.ContainsKey(sn))
{
continue; // такой ключ уже был
}
exists.Add(sn, 0);
int sidx = indexes[sn];
for (int j = 0; j < modNvars; j++)
{
nxy[i, j] = xy[sidx, vidxes[j]];
}
nxy[i, modNvars] = xy[sidx, nvars];
nk++;
if (nk >= modNpoints)
{
break;
}
}
if (nk >= modNpoints)
{
break;
}
}
}
var ridxes = exists.Keys.ToArray();
return(CreateTree(nxy, nclasses, nvars, modNpoints, modNvars, vidxes, ridxes));
}
