BinaryClassificationCatalog 類別

定義

用來 MLContext 建立二元分類元件的實例的類別,例如定型器和校正器。

public sealed class BinaryClassificationCatalog : Microsoft.ML.TrainCatalogBase
type BinaryClassificationCatalog = class
    inherit TrainCatalogBase
Public NotInheritable Class BinaryClassificationCatalog
Inherits TrainCatalogBase
繼承
BinaryClassificationCatalog

屬性

Calibrators

用於執行二元分類的校正器清單。

Trainers

執行二元分類的定型器清單。

方法

ChangeModelThreshold<TModel>(BinaryPredictionTransformer<TModel>, Single)

將臨界值修改為現有模型並傳回修改過的模型的方法。

CrossValidate(IDataView, IEstimator<ITransformer>, Int32, String, String, Nullable<Int32>)

在的折迭上numberOfFolds執行交叉驗證,方法是調整 estimator,並在提供時遵守 samplingKeyColumnNamedata 然後針對 labelColumnName 每個子模型評估每個子模型,並傳回 CalibratedBinaryClassificationMetrics 物件,其中包含每個子模型的機率型計量。 每個子模型都會在定型期間看不到的交叉驗證折疊上進行評估。

CrossValidateNonCalibrated(IDataView, IEstimator<ITransformer>, Int32, String, String, Nullable<Int32>)

在的折迭上numberOfFolds執行交叉驗證,方法是調整 estimator,並在提供時遵守 samplingKeyColumnNamedata 然後針對 labelColumnName 每個子模型評估每個子模型,並傳回 BinaryClassificationMetrics 不包含每個子模型的機率型計量的物件。 每個子模型都會在定型期間看不到的交叉驗證折疊上進行評估。

Evaluate(IDataView, String, String, String, String)

評估評分的二進位分類數據。

EvaluateNonCalibrated(IDataView, String, String, String)

評估評分的二進位分類數據,而不需以機率為基礎的計量。

擴充方法

PermutationFeatureImportance<TModel>(BinaryClassificationCatalog, ISingleFeaturePredictionTransformer<TModel>, IDataView, String, Boolean, Nullable<Int32>, Int32)

二元分類的 PFI) (排列特徵重要性。

PermutationFeatureImportanceNonCalibrated(BinaryClassificationCatalog, ITransformer, IDataView, String, Boolean, Nullable<Int32>, Int32)

二元分類的 PFI) (排列特徵重要性。

適用於