2026-05-20 21:25:14 +03:00
|
|
|
|
// Confusion matrix для Iris с использованием DataPipeline.
|
|
|
|
|
|
// Рекомендуемый стиль - он проще и явно показывает намерения
|
|
|
|
|
|
|
|
|
|
|
|
uses MLABC, PlotML;
|
|
|
|
|
|
|
|
|
|
|
|
begin
|
|
|
|
|
|
var ds := Datasets.Iris;
|
|
|
|
|
|
var (train, test) := ds.StratifiedTrainTestSplit(0.2, 42);
|
|
|
|
|
|
|
2026-05-28 10:23:22 +03:00
|
|
|
|
var pipe := DataPipeline.BuildClassification(
|
2026-05-20 21:25:14 +03:00
|
|
|
|
ds.Target,
|
|
|
|
|
|
ds.Features,
|
|
|
|
|
|
new StandardScaler,
|
|
|
|
|
|
new LogisticRegression
|
|
|
|
|
|
);
|
|
|
|
|
|
|
|
|
|
|
|
pipe.Fit(train.Data);
|
|
|
|
|
|
|
|
|
|
|
|
var pred := pipe.Predict(test.Data);
|
|
|
|
|
|
var ytest := pipe.GetEncodedLabels(test.Data);
|
|
|
|
|
|
var cm := new ConfusionMatrix(ytest, pred);
|
|
|
|
|
|
var acc := Metrics.Accuracy(ytest, pred);
|
|
|
|
|
|
|
|
|
|
|
|
Plot.ConfusionMatrix(cm, pipe.GetClassLabels{, normalize := MatrixNormalization.Rows});
|
|
|
|
|
|
Plot.Title := $'Iris: DataPipeline, accuracy = {acc:F3}';
|
|
|
|
|
|
Plot.XLabel := 'Предсказанный класс';
|
|
|
|
|
|
Plot.YLabel := 'Истинный класс';
|
|
|
|
|
|
end.
|
