7eaddd9a54
ML - устранение неточностей и багов ML - оптимизация производительности DecisionTreeRegressor.Fit, RandomForestRegressor.Fit ML - тесты
122 lines
2.6 KiB
ObjectPascal
122 lines
2.6 KiB
ObjectPascal
uses MLABC, PlotML;
|
|
|
|
function ToIntArray(v: Vector): array of integer;
|
|
begin
|
|
Result := v.Data.Select(t -> integer(t)).ToArray;
|
|
end;
|
|
|
|
function BinLabels(v: Vector; bins: integer := 8): array of integer;
|
|
begin
|
|
Result := new integer[v.Length];
|
|
|
|
var vmin := v.Min;
|
|
var vmax := v.Max;
|
|
|
|
if vmax = vmin then
|
|
exit;
|
|
|
|
var w := (vmax - vmin) / bins;
|
|
|
|
for var i := 0 to v.Length - 1 do
|
|
begin
|
|
var k := trunc((v[i] - vmin) / w);
|
|
if k >= bins then k := bins - 1;
|
|
if k < 0 then k := 0;
|
|
Result[i] := k;
|
|
end;
|
|
end;
|
|
|
|
procedure DrawDataset(cell: Cell; X: Matrix; labels: array of integer; title: string);
|
|
begin
|
|
var x1 := X.Col(0);
|
|
var x2 := X.Col(1);
|
|
|
|
// защита от рассинхронизации
|
|
if labels.Length <> X.RowCount then
|
|
raise new Exception('Labels length mismatch');
|
|
|
|
cell.SetPalette(Palettes.Bright);
|
|
cell.Points(x1, x2, labels, size := 5);
|
|
cell.Title := title;
|
|
end;
|
|
|
|
begin
|
|
var fig := Plot.Grid(2, 3);
|
|
|
|
// --- Blobs
|
|
var (X1, y1) := Datasets.MakeBlobs(
|
|
n := 300,
|
|
centers := 3,
|
|
nFeatures := 2,
|
|
clusterStd := 0.7,
|
|
clusterStdVar := 0.4,
|
|
centerBox := 5.0,
|
|
classBalance := 1.0,
|
|
noisePoints := 20,
|
|
shuffle := True,
|
|
seed := 1
|
|
);
|
|
DrawDataset(fig[0,0], X1, ToIntArray(y1), 'MakeBlobs');
|
|
|
|
// --- Moons
|
|
var (X2, y2) := Datasets.MakeMoons(
|
|
n := 300,
|
|
noise := 0.1,
|
|
shuffle := True,
|
|
seed := 2
|
|
);
|
|
DrawDataset(fig[0,1], X2, ToIntArray(y2), 'MakeMoons');
|
|
|
|
// --- Circles
|
|
var (X3, y3) := Datasets.MakeCircles(
|
|
n := 300,
|
|
noise := 0.08,
|
|
factor := 0.45,
|
|
classBalance := 0.5,
|
|
flipProb := 0.0,
|
|
scale := 3.0,
|
|
shuffle := True,
|
|
seed := 3
|
|
);
|
|
DrawDataset(fig[0,2], X3, ToIntArray(y3), 'MakeCircles');
|
|
|
|
// --- Spiral
|
|
var (X4, y4) := Datasets.MakeSpiral(
|
|
n := 300,
|
|
noise := 0.03,
|
|
turns := 2.5,
|
|
shuffle := True,
|
|
seed := 4
|
|
);
|
|
DrawDataset(fig[1,0], X4, ToIntArray(y4), 'MakeSpiral');
|
|
|
|
// --- Regression (биннинг)
|
|
var (X5, y5) := Datasets.MakeRegression(
|
|
n := 300,
|
|
nFeatures := 2,
|
|
nInformative := 2,
|
|
noise := 0.02,
|
|
coefScale := 1.0,
|
|
bias := 0.0,
|
|
nonlinearStrength := 3.0,
|
|
shuffle := True,
|
|
seed := 5
|
|
);
|
|
var labels := ArrFill(X5.RowCount, 0);
|
|
DrawDataset(fig[1,1], X5, labels, 'MakeRegression');
|
|
|
|
// --- Classification
|
|
var (X6, y6) := Datasets.MakeClassification(
|
|
n := 300,
|
|
nFeatures := 2,
|
|
nInformative := 2,
|
|
nRedundant := 0,
|
|
noise := 0.2,
|
|
classSep := 2.5,
|
|
flipProb := 0.05,
|
|
classBalance := 0.5,
|
|
shuffle := True,
|
|
seed := 6
|
|
);
|
|
DrawDataset(fig[1,2], X6, ToIntArray(y6), 'MakeClassification');
|
|
end. |