BinaryClassificationCatalog 類別

定義

類別 用於 MLContext 建立二元分類元件的實例,例如訓練器與校準器。

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

屬性

名稱 Description
Calibrators

執行二元分類的校準器列表。

Trainers

執行二元分類的訓練師名單。

方法

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

修改閾值至現有模型並回傳修改後模型的方法。

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

對 的摺numberOfFolds疊進行交叉驗證data,方法是擬estimator合 ,若提供 則尊重 samplingKeyColumnName 。 接著對每個子模型 labelColumnName 進行評估,並回傳一個 CalibratedBinaryClassificationMetrics 包含機率基礎指標的物件。 每個子模型會根據其在訓練時未察覺的交叉驗證摺疊進行評估。

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

對 的摺numberOfFolds疊進行交叉驗證data,方法是擬estimator合 ,若提供 則尊重 samplingKeyColumnName 。 接著對每個子模型 labelColumnName 進行評估,並回傳一個 BinaryClassificationMetrics 不包含機率度量的物件,針對每個子模型。 每個子模型會根據其在訓練時未察覺的交叉驗證摺疊進行評估。

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

評估評分過的二元分類資料。

EvaluateNonCalibrated(IDataView, String, String, String)

評估評分的二元分類資料,不使用機率基礎指標。

擴充方法

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

二元分類中的置換特徵重要性(PFI)。

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

二元分類中的置換特徵重要性(PFI)。

適用於